Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_nl.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
b411b363 PR |
26 | #include <linux/module.h> |
27 | #include <linux/drbd.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/fs.h> | |
30 | #include <linux/file.h> | |
31 | #include <linux/slab.h> | |
b411b363 PR |
32 | #include <linux/blkpg.h> |
33 | #include <linux/cpumask.h> | |
34 | #include "drbd_int.h" | |
265be2d0 | 35 | #include "drbd_req.h" |
b411b363 PR |
36 | #include "drbd_wrappers.h" |
37 | #include <asm/unaligned.h> | |
b411b363 | 38 | #include <linux/drbd_limits.h> |
87f7be4c | 39 | #include <linux/kthread.h> |
b411b363 | 40 | |
3b98c0c2 LE |
41 | #include <net/genetlink.h> |
42 | ||
43 | /* .doit */ | |
44 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
45 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | ||
47 | int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info); | |
48 | int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info); | |
49 | ||
50 | int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info); | |
51 | int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 52 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
53 | |
54 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
55 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 56 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
57 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
58 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
62 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 71 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
72 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
73 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
74 | /* .dumpit */ | |
75 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
76 | ||
77 | #include <linux/drbd_genl_api.h> | |
78 | #include <linux/genl_magic_func.h> | |
79 | ||
80 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
81 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
82 | ||
3b98c0c2 LE |
83 | /* Configuration is strictly serialized, because generic netlink message |
84 | * processing is strictly serialized by the genl_lock(). | |
85 | * Which means we can use one static global drbd_config_context struct. | |
86 | */ | |
87 | static struct drbd_config_context { | |
88 | /* assigned from drbd_genlmsghdr */ | |
89 | unsigned int minor; | |
90 | /* assigned from request attributes, if present */ | |
91 | unsigned int volume; | |
92 | #define VOLUME_UNSPECIFIED (-1U) | |
93 | /* pointer into the request skb, | |
94 | * limited lifetime! */ | |
95 | char *conn_name; | |
96 | ||
97 | /* reply buffer */ | |
98 | struct sk_buff *reply_skb; | |
99 | /* pointer into reply buffer */ | |
100 | struct drbd_genlmsghdr *reply_dh; | |
101 | /* resolved from attributes, if possible */ | |
102 | struct drbd_conf *mdev; | |
103 | struct drbd_tconn *tconn; | |
104 | } adm_ctx; | |
105 | ||
106 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) | |
107 | { | |
108 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
109 | if (genlmsg_reply(skb, info)) | |
110 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 111 | } |
3b98c0c2 LE |
112 | |
113 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
114 | * reason it could fail was no space in skb, and there are 4k available. */ | |
8432b314 | 115 | int drbd_msg_put_info(const char *info) |
3b98c0c2 LE |
116 | { |
117 | struct sk_buff *skb = adm_ctx.reply_skb; | |
118 | struct nlattr *nla; | |
119 | int err = -EMSGSIZE; | |
120 | ||
121 | if (!info || !info[0]) | |
122 | return 0; | |
123 | ||
124 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
125 | if (!nla) | |
126 | return err; | |
127 | ||
128 | err = nla_put_string(skb, T_info_text, info); | |
129 | if (err) { | |
130 | nla_nest_cancel(skb, nla); | |
131 | return err; | |
132 | } else | |
133 | nla_nest_end(skb, nla); | |
134 | return 0; | |
b411b363 PR |
135 | } |
136 | ||
3b98c0c2 LE |
137 | /* This would be a good candidate for a "pre_doit" hook, |
138 | * and per-family private info->pointers. | |
139 | * But we need to stay compatible with older kernels. | |
140 | * If it returns successfully, adm_ctx members are valid. | |
141 | */ | |
142 | #define DRBD_ADM_NEED_MINOR 1 | |
143 | #define DRBD_ADM_NEED_CONN 2 | |
144 | static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info, | |
145 | unsigned flags) | |
146 | { | |
147 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
148 | const u8 cmd = info->genlhdr->cmd; | |
149 | int err; | |
150 | ||
151 | memset(&adm_ctx, 0, sizeof(adm_ctx)); | |
152 | ||
153 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
154 | if (cmd != DRBD_ADM_GET_STATUS | |
155 | && security_netlink_recv(skb, CAP_SYS_ADMIN)) | |
156 | return -EPERM; | |
157 | ||
158 | adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); | |
159 | if (!adm_ctx.reply_skb) | |
160 | goto fail; | |
161 | ||
162 | adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb, | |
163 | info, &drbd_genl_family, 0, cmd); | |
164 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
165 | * but anyways */ | |
166 | if (!adm_ctx.reply_dh) | |
167 | goto fail; | |
168 | ||
169 | adm_ctx.reply_dh->minor = d_in->minor; | |
170 | adm_ctx.reply_dh->ret_code = NO_ERROR; | |
171 | ||
172 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { | |
173 | struct nlattr *nla; | |
174 | /* parse and validate only */ | |
f399002e | 175 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
176 | if (err) |
177 | goto fail; | |
178 | ||
179 | /* It was present, and valid, | |
180 | * copy it over to the reply skb. */ | |
181 | err = nla_put_nohdr(adm_ctx.reply_skb, | |
182 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, | |
183 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
184 | if (err) | |
185 | goto fail; | |
186 | ||
187 | /* and assign stuff to the global adm_ctx */ | |
188 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; | |
189 | adm_ctx.volume = nla ? nla_get_u32(nla) : VOLUME_UNSPECIFIED; | |
190 | nla = nested_attr_tb[__nla_type(T_ctx_conn_name)]; | |
191 | if (nla) | |
192 | adm_ctx.conn_name = nla_data(nla); | |
193 | } else | |
194 | adm_ctx.volume = VOLUME_UNSPECIFIED; | |
195 | ||
196 | adm_ctx.minor = d_in->minor; | |
197 | adm_ctx.mdev = minor_to_mdev(d_in->minor); | |
0ace9dfa | 198 | adm_ctx.tconn = conn_get_by_name(adm_ctx.conn_name); |
3b98c0c2 LE |
199 | |
200 | if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) { | |
201 | drbd_msg_put_info("unknown minor"); | |
202 | return ERR_MINOR_INVALID; | |
203 | } | |
204 | if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_CONN)) { | |
205 | drbd_msg_put_info("unknown connection"); | |
206 | return ERR_INVALID_REQUEST; | |
207 | } | |
208 | ||
209 | /* some more paranoia, if the request was over-determined */ | |
527f4b24 LE |
210 | if (adm_ctx.mdev && adm_ctx.tconn && |
211 | adm_ctx.mdev->tconn != adm_ctx.tconn) { | |
212 | pr_warning("request: minor=%u, conn=%s; but that minor belongs to connection %s\n", | |
213 | adm_ctx.minor, adm_ctx.conn_name, adm_ctx.mdev->tconn->name); | |
214 | drbd_msg_put_info("minor exists in different connection"); | |
215 | return ERR_INVALID_REQUEST; | |
216 | } | |
3b98c0c2 LE |
217 | if (adm_ctx.mdev && |
218 | adm_ctx.volume != VOLUME_UNSPECIFIED && | |
219 | adm_ctx.volume != adm_ctx.mdev->vnr) { | |
220 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", | |
221 | adm_ctx.minor, adm_ctx.volume, | |
222 | adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->name); | |
527f4b24 | 223 | drbd_msg_put_info("minor exists as different volume"); |
3b98c0c2 LE |
224 | return ERR_INVALID_REQUEST; |
225 | } | |
0ace9dfa | 226 | |
3b98c0c2 LE |
227 | return NO_ERROR; |
228 | ||
229 | fail: | |
230 | nlmsg_free(adm_ctx.reply_skb); | |
231 | adm_ctx.reply_skb = NULL; | |
232 | return -ENOMEM; | |
233 | } | |
234 | ||
235 | static int drbd_adm_finish(struct genl_info *info, int retcode) | |
236 | { | |
237 | struct nlattr *nla; | |
238 | const char *conn_name = NULL; | |
239 | ||
0ace9dfa PR |
240 | if (adm_ctx.tconn) { |
241 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
242 | adm_ctx.tconn = NULL; | |
243 | } | |
244 | ||
3b98c0c2 LE |
245 | if (!adm_ctx.reply_skb) |
246 | return -ENOMEM; | |
247 | ||
248 | adm_ctx.reply_dh->ret_code = retcode; | |
249 | ||
250 | nla = info->attrs[DRBD_NLA_CFG_CONTEXT]; | |
251 | if (nla) { | |
252 | nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name)); | |
253 | if (nla) | |
254 | conn_name = nla_data(nla); | |
255 | } | |
256 | ||
257 | drbd_adm_send_reply(adm_ctx.reply_skb, info); | |
258 | return 0; | |
259 | } | |
b411b363 | 260 | |
6b75dced | 261 | static void setup_khelper_env(struct drbd_tconn *tconn, char **envp) |
b411b363 | 262 | { |
6b75dced | 263 | char *afs; |
44ed167d | 264 | struct net_conf *nc; |
b411b363 | 265 | |
44ed167d PR |
266 | rcu_read_lock(); |
267 | nc = rcu_dereference(tconn->net_conf); | |
268 | if (nc) { | |
269 | switch (((struct sockaddr *)nc->peer_addr)->sa_family) { | |
b411b363 PR |
270 | case AF_INET6: |
271 | afs = "ipv6"; | |
6b75dced | 272 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", |
44ed167d | 273 | &((struct sockaddr_in6 *)nc->peer_addr)->sin6_addr); |
b411b363 PR |
274 | break; |
275 | case AF_INET: | |
276 | afs = "ipv4"; | |
6b75dced | 277 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", |
44ed167d | 278 | &((struct sockaddr_in *)nc->peer_addr)->sin_addr); |
b411b363 PR |
279 | break; |
280 | default: | |
281 | afs = "ssocks"; | |
6b75dced | 282 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", |
44ed167d | 283 | &((struct sockaddr_in *)nc->peer_addr)->sin_addr); |
b411b363 | 284 | } |
6b75dced | 285 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
b411b363 | 286 | } |
44ed167d | 287 | rcu_read_unlock(); |
6b75dced PR |
288 | } |
289 | ||
290 | int drbd_khelper(struct drbd_conf *mdev, char *cmd) | |
291 | { | |
292 | char *envp[] = { "HOME=/", | |
293 | "TERM=linux", | |
294 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
295 | (char[20]) { }, /* address family */ | |
296 | (char[60]) { }, /* address */ | |
297 | NULL }; | |
298 | char mb[12]; | |
299 | char *argv[] = {usermode_helper, cmd, mb, NULL }; | |
300 | struct sib_info sib; | |
301 | int ret; | |
302 | ||
303 | snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev)); | |
304 | setup_khelper_env(mdev->tconn, envp); | |
b411b363 | 305 | |
1090c056 LE |
306 | /* The helper may take some time. |
307 | * write out any unsynced meta data changes now */ | |
308 | drbd_md_sync(mdev); | |
309 | ||
b411b363 | 310 | dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
311 | sib.sib_reason = SIB_HELPER_PRE; |
312 | sib.helper_name = cmd; | |
313 | drbd_bcast_event(mdev, &sib); | |
b411b363 PR |
314 | ret = call_usermodehelper(usermode_helper, argv, envp, 1); |
315 | if (ret) | |
316 | dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
317 | usermode_helper, cmd, mb, | |
318 | (ret >> 8) & 0xff, ret); | |
319 | else | |
320 | dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
321 | usermode_helper, cmd, mb, | |
322 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
323 | sib.sib_reason = SIB_HELPER_POST; |
324 | sib.helper_exit_code = ret; | |
325 | drbd_bcast_event(mdev, &sib); | |
b411b363 PR |
326 | |
327 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
328 | ret = 0; | |
329 | ||
330 | return ret; | |
331 | } | |
332 | ||
6b75dced PR |
333 | static void conn_md_sync(struct drbd_tconn *tconn) |
334 | { | |
335 | struct drbd_conf *mdev; | |
e90285e0 | 336 | int vnr; |
6b75dced | 337 | |
d3fcb490 | 338 | down_read(&drbd_cfg_rwsem); |
e90285e0 | 339 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
6b75dced | 340 | drbd_md_sync(mdev); |
d3fcb490 | 341 | up_read(&drbd_cfg_rwsem); |
6b75dced PR |
342 | } |
343 | ||
344 | int conn_khelper(struct drbd_tconn *tconn, char *cmd) | |
345 | { | |
346 | char *envp[] = { "HOME=/", | |
347 | "TERM=linux", | |
348 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
349 | (char[20]) { }, /* address family */ | |
350 | (char[60]) { }, /* address */ | |
351 | NULL }; | |
352 | char *argv[] = {usermode_helper, cmd, tconn->name, NULL }; | |
353 | int ret; | |
354 | ||
355 | setup_khelper_env(tconn, envp); | |
356 | conn_md_sync(tconn); | |
357 | ||
358 | conn_info(tconn, "helper command: %s %s %s\n", usermode_helper, cmd, tconn->name); | |
359 | /* TODO: conn_bcast_event() ?? */ | |
360 | ||
361 | ret = call_usermodehelper(usermode_helper, argv, envp, 1); | |
362 | if (ret) | |
363 | conn_warn(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
364 | usermode_helper, cmd, tconn->name, | |
365 | (ret >> 8) & 0xff, ret); | |
366 | else | |
367 | conn_info(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
368 | usermode_helper, cmd, tconn->name, | |
369 | (ret >> 8) & 0xff, ret); | |
370 | /* TODO: conn_bcast_event() ?? */ | |
371 | ||
372 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
373 | ret = 0; | |
374 | ||
375 | return ret; | |
376 | } | |
377 | ||
cb703454 | 378 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn) |
b411b363 | 379 | { |
cb703454 PR |
380 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
381 | struct drbd_conf *mdev; | |
382 | int vnr; | |
383 | ||
695d08fa | 384 | rcu_read_lock(); |
cb703454 PR |
385 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
386 | if (get_ldev_if_state(mdev, D_CONSISTENT)) { | |
387 | fp = max_t(enum drbd_fencing_p, fp, mdev->ldev->dc.fencing); | |
388 | put_ldev(mdev); | |
389 | } | |
390 | } | |
695d08fa | 391 | rcu_read_unlock(); |
cb703454 PR |
392 | |
393 | return fp; | |
394 | } | |
395 | ||
396 | bool conn_try_outdate_peer(struct drbd_tconn *tconn) | |
397 | { | |
398 | union drbd_state mask = { }; | |
399 | union drbd_state val = { }; | |
400 | enum drbd_fencing_p fp; | |
b411b363 PR |
401 | char *ex_to_string; |
402 | int r; | |
b411b363 | 403 | |
cb703454 PR |
404 | if (tconn->cstate >= C_WF_REPORT_PARAMS) { |
405 | conn_err(tconn, "Expected cstate < C_WF_REPORT_PARAMS\n"); | |
406 | return false; | |
407 | } | |
b411b363 | 408 | |
cb703454 PR |
409 | fp = highest_fencing_policy(tconn); |
410 | switch (fp) { | |
411 | case FP_NOT_AVAIL: | |
412 | conn_warn(tconn, "Not fencing peer, I'm not even Consistent myself.\n"); | |
fb22c402 | 413 | goto out; |
cb703454 PR |
414 | case FP_DONT_CARE: |
415 | return true; | |
416 | default: ; | |
b411b363 PR |
417 | } |
418 | ||
cb703454 | 419 | r = conn_khelper(tconn, "fence-peer"); |
b411b363 PR |
420 | |
421 | switch ((r>>8) & 0xff) { | |
422 | case 3: /* peer is inconsistent */ | |
423 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
424 | mask.pdsk = D_MASK; |
425 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
426 | break; |
427 | case 4: /* peer got outdated, or was already outdated */ | |
428 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
429 | mask.pdsk = D_MASK; |
430 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
431 | break; |
432 | case 5: /* peer was down */ | |
cb703454 | 433 | if (conn_highest_disk(tconn) == D_UP_TO_DATE) { |
b411b363 PR |
434 | /* we will(have) create(d) a new UUID anyways... */ |
435 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
436 | mask.pdsk = D_MASK; |
437 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
438 | } else { |
439 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
440 | } |
441 | break; | |
442 | case 6: /* Peer is primary, voluntarily outdate myself. | |
443 | * This is useful when an unconnected R_SECONDARY is asked to | |
444 | * become R_PRIMARY, but finds the other peer being active. */ | |
445 | ex_to_string = "peer is active"; | |
cb703454 PR |
446 | conn_warn(tconn, "Peer is primary, outdating myself.\n"); |
447 | mask.disk = D_MASK; | |
448 | val.disk = D_OUTDATED; | |
b411b363 PR |
449 | break; |
450 | case 7: | |
451 | if (fp != FP_STONITH) | |
cb703454 | 452 | conn_err(tconn, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 453 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
454 | mask.pdsk = D_MASK; |
455 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
456 | break; |
457 | default: | |
458 | /* The script is broken ... */ | |
cb703454 PR |
459 | conn_err(tconn, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
460 | return false; /* Eventually leave IO frozen */ | |
b411b363 PR |
461 | } |
462 | ||
cb703454 PR |
463 | conn_info(tconn, "fence-peer helper returned %d (%s)\n", |
464 | (r>>8) & 0xff, ex_to_string); | |
fb22c402 | 465 | |
cb703454 | 466 | out: |
fb22c402 | 467 | |
cb703454 PR |
468 | /* Not using |
469 | conn_request_state(tconn, mask, val, CS_VERBOSE); | |
470 | here, because we might were able to re-establish the connection in the | |
471 | meantime. */ | |
472 | spin_lock_irq(&tconn->req_lock); | |
473 | if (tconn->cstate < C_WF_REPORT_PARAMS) | |
474 | _conn_request_state(tconn, mask, val, CS_VERBOSE); | |
475 | spin_unlock_irq(&tconn->req_lock); | |
476 | ||
477 | return conn_highest_pdsk(tconn) <= D_OUTDATED; | |
b411b363 PR |
478 | } |
479 | ||
87f7be4c PR |
480 | static int _try_outdate_peer_async(void *data) |
481 | { | |
cb703454 | 482 | struct drbd_tconn *tconn = (struct drbd_tconn *)data; |
87f7be4c | 483 | |
cb703454 | 484 | conn_try_outdate_peer(tconn); |
87f7be4c | 485 | |
9dc9fbb3 | 486 | kref_put(&tconn->kref, &conn_destroy); |
87f7be4c PR |
487 | return 0; |
488 | } | |
489 | ||
cb703454 | 490 | void conn_try_outdate_peer_async(struct drbd_tconn *tconn) |
87f7be4c PR |
491 | { |
492 | struct task_struct *opa; | |
493 | ||
9dc9fbb3 | 494 | kref_get(&tconn->kref); |
cb703454 | 495 | opa = kthread_run(_try_outdate_peer_async, tconn, "drbd_async_h"); |
9dc9fbb3 | 496 | if (IS_ERR(opa)) { |
cb703454 | 497 | conn_err(tconn, "out of mem, failed to invoke fence-peer helper\n"); |
9dc9fbb3 PR |
498 | kref_put(&tconn->kref, &conn_destroy); |
499 | } | |
87f7be4c | 500 | } |
b411b363 | 501 | |
bf885f8a AG |
502 | enum drbd_state_rv |
503 | drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) | |
b411b363 PR |
504 | { |
505 | const int max_tries = 4; | |
bf885f8a | 506 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 507 | struct net_conf *nc; |
b411b363 PR |
508 | int try = 0; |
509 | int forced = 0; | |
510 | union drbd_state mask, val; | |
b411b363 PR |
511 | |
512 | if (new_role == R_PRIMARY) | |
0625ac19 | 513 | request_ping(mdev->tconn); /* Detect a dead peer ASAP */ |
b411b363 | 514 | |
8410da8f | 515 | mutex_lock(mdev->state_mutex); |
b411b363 PR |
516 | |
517 | mask.i = 0; mask.role = R_MASK; | |
518 | val.i = 0; val.role = new_role; | |
519 | ||
520 | while (try++ < max_tries) { | |
bf885f8a | 521 | rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
522 | |
523 | /* in case we first succeeded to outdate, | |
524 | * but now suddenly could establish a connection */ | |
bf885f8a | 525 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
526 | val.pdsk = 0; |
527 | mask.pdsk = 0; | |
528 | continue; | |
529 | } | |
530 | ||
bf885f8a | 531 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
d10a33c6 PR |
532 | (mdev->state.disk < D_UP_TO_DATE && |
533 | mdev->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
534 | mask.disk = D_MASK; |
535 | val.disk = D_UP_TO_DATE; | |
536 | forced = 1; | |
537 | continue; | |
538 | } | |
539 | ||
bf885f8a | 540 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b411b363 PR |
541 | mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
542 | D_ASSERT(mdev->state.pdsk == D_UNKNOWN); | |
b411b363 | 543 | |
cb703454 | 544 | if (conn_try_outdate_peer(mdev->tconn)) { |
b411b363 PR |
545 | val.disk = D_UP_TO_DATE; |
546 | mask.disk = D_MASK; | |
547 | } | |
b411b363 PR |
548 | continue; |
549 | } | |
550 | ||
bf885f8a | 551 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 552 | goto out; |
bf885f8a | 553 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
cb703454 | 554 | if (!conn_try_outdate_peer(mdev->tconn) && force) { |
b411b363 | 555 | dev_warn(DEV, "Forced into split brain situation!\n"); |
cb703454 PR |
556 | mask.pdsk = D_MASK; |
557 | val.pdsk = D_OUTDATED; | |
b411b363 | 558 | |
cb703454 | 559 | } |
b411b363 PR |
560 | continue; |
561 | } | |
bf885f8a | 562 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
563 | /* Maybe the peer is detected as dead very soon... |
564 | retry at most once more in this case. */ | |
44ed167d PR |
565 | int timeo; |
566 | rcu_read_lock(); | |
567 | nc = rcu_dereference(mdev->tconn->net_conf); | |
568 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; | |
569 | rcu_read_unlock(); | |
570 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
571 | if (try < max_tries) |
572 | try = max_tries - 1; | |
573 | continue; | |
574 | } | |
bf885f8a AG |
575 | if (rv < SS_SUCCESS) { |
576 | rv = _drbd_request_state(mdev, mask, val, | |
b411b363 | 577 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 578 | if (rv < SS_SUCCESS) |
3b98c0c2 | 579 | goto out; |
b411b363 PR |
580 | } |
581 | break; | |
582 | } | |
583 | ||
bf885f8a | 584 | if (rv < SS_SUCCESS) |
3b98c0c2 | 585 | goto out; |
b411b363 PR |
586 | |
587 | if (forced) | |
588 | dev_warn(DEV, "Forced to consider local data as UpToDate!\n"); | |
589 | ||
590 | /* Wait until nothing is on the fly :) */ | |
591 | wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0); | |
592 | ||
593 | if (new_role == R_SECONDARY) { | |
81e84650 | 594 | set_disk_ro(mdev->vdisk, true); |
b411b363 PR |
595 | if (get_ldev(mdev)) { |
596 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
597 | put_ldev(mdev); | |
598 | } | |
599 | } else { | |
91fd4dad PR |
600 | mutex_lock(&mdev->tconn->net_conf_update); |
601 | nc = mdev->tconn->net_conf; | |
44ed167d | 602 | if (nc) |
91fd4dad PR |
603 | nc->want_lose = 0; /* without copy; single bit op is atomic */ |
604 | mutex_unlock(&mdev->tconn->net_conf_update); | |
605 | ||
81e84650 | 606 | set_disk_ro(mdev->vdisk, false); |
b411b363 PR |
607 | if (get_ldev(mdev)) { |
608 | if (((mdev->state.conn < C_CONNECTED || | |
609 | mdev->state.pdsk <= D_FAILED) | |
610 | && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
611 | drbd_uuid_new_current(mdev); | |
612 | ||
613 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
614 | put_ldev(mdev); | |
615 | } | |
616 | } | |
617 | ||
19f843aa LE |
618 | /* writeout of activity log covered areas of the bitmap |
619 | * to stable storage done in after state change already */ | |
b411b363 PR |
620 | |
621 | if (mdev->state.conn >= C_WF_REPORT_PARAMS) { | |
622 | /* if this was forced, we should consider sync */ | |
623 | if (forced) | |
624 | drbd_send_uuids(mdev); | |
625 | drbd_send_state(mdev); | |
626 | } | |
627 | ||
628 | drbd_md_sync(mdev); | |
629 | ||
630 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
3b98c0c2 | 631 | out: |
8410da8f | 632 | mutex_unlock(mdev->state_mutex); |
bf885f8a | 633 | return rv; |
b411b363 PR |
634 | } |
635 | ||
3b98c0c2 | 636 | static const char *from_attrs_err_to_txt(int err) |
b411b363 | 637 | { |
3b98c0c2 LE |
638 | return err == -ENOMSG ? "required attribute missing" : |
639 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 640 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 641 | "invalid attribute value"; |
b411b363 PR |
642 | } |
643 | ||
3b98c0c2 | 644 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 645 | { |
3b98c0c2 LE |
646 | struct set_role_parms parms; |
647 | int err; | |
648 | enum drbd_ret_code retcode; | |
b411b363 | 649 | |
3b98c0c2 LE |
650 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
651 | if (!adm_ctx.reply_skb) | |
652 | return retcode; | |
653 | if (retcode != NO_ERROR) | |
654 | goto out; | |
655 | ||
656 | memset(&parms, 0, sizeof(parms)); | |
657 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 658 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
659 | if (err) { |
660 | retcode = ERR_MANDATORY_TAG; | |
661 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
662 | goto out; | |
663 | } | |
664 | } | |
665 | ||
666 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) | |
667 | retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate); | |
668 | else | |
669 | retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0); | |
670 | out: | |
671 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
672 | return 0; |
673 | } | |
674 | ||
675 | /* initializes the md.*_offset members, so we are able to find | |
676 | * the on disk meta data */ | |
677 | static void drbd_md_set_sector_offsets(struct drbd_conf *mdev, | |
678 | struct drbd_backing_dev *bdev) | |
679 | { | |
680 | sector_t md_size_sect = 0; | |
681 | switch (bdev->dc.meta_dev_idx) { | |
682 | default: | |
683 | /* v07 style fixed size indexed meta data */ | |
684 | bdev->md.md_size_sect = MD_RESERVED_SECT; | |
685 | bdev->md.md_offset = drbd_md_ss__(mdev, bdev); | |
686 | bdev->md.al_offset = MD_AL_OFFSET; | |
687 | bdev->md.bm_offset = MD_BM_OFFSET; | |
688 | break; | |
689 | case DRBD_MD_INDEX_FLEX_EXT: | |
690 | /* just occupy the full device; unit: sectors */ | |
691 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
692 | bdev->md.md_offset = 0; | |
693 | bdev->md.al_offset = MD_AL_OFFSET; | |
694 | bdev->md.bm_offset = MD_BM_OFFSET; | |
695 | break; | |
696 | case DRBD_MD_INDEX_INTERNAL: | |
697 | case DRBD_MD_INDEX_FLEX_INT: | |
698 | bdev->md.md_offset = drbd_md_ss__(mdev, bdev); | |
699 | /* al size is still fixed */ | |
7ad651b5 | 700 | bdev->md.al_offset = -MD_AL_SECTORS; |
b411b363 PR |
701 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
702 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
703 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
704 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
705 | md_size_sect = ALIGN(md_size_sect, 8); | |
706 | ||
707 | /* plus the "drbd meta data super block", | |
708 | * and the activity log; */ | |
709 | md_size_sect += MD_BM_OFFSET; | |
710 | ||
711 | bdev->md.md_size_sect = md_size_sect; | |
712 | /* bitmap offset is adjusted by 'super' block size */ | |
713 | bdev->md.bm_offset = -md_size_sect + MD_AL_OFFSET; | |
714 | break; | |
715 | } | |
716 | } | |
717 | ||
4b0715f0 | 718 | /* input size is expected to be in KB */ |
b411b363 PR |
719 | char *ppsize(char *buf, unsigned long long size) |
720 | { | |
4b0715f0 LE |
721 | /* Needs 9 bytes at max including trailing NUL: |
722 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
723 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
724 | int base = 0; | |
4b0715f0 | 725 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
726 | /* shift + round */ |
727 | size = (size >> 10) + !!(size & (1<<9)); | |
728 | base++; | |
729 | } | |
4b0715f0 | 730 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
731 | |
732 | return buf; | |
733 | } | |
734 | ||
735 | /* there is still a theoretical deadlock when called from receiver | |
736 | * on an D_INCONSISTENT R_PRIMARY: | |
737 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
738 | * packet from remote to dec_ap_bio again. | |
739 | * receiver receive_sizes(), comes here, | |
740 | * waits for ap_bio_cnt == 0. -> deadlock. | |
741 | * but this cannot happen, actually, because: | |
742 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
743 | * (not connected, or bad/no disk on peer): | |
744 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
745 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
746 | * peer may not initiate a resize. | |
747 | */ | |
3b98c0c2 LE |
748 | /* Note these are not to be confused with |
749 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
750 | * which are (sub) state changes triggered by admin (drbdsetup), | |
751 | * and can be long lived. | |
752 | * This changes an mdev->flag, is triggered by drbd internals, | |
753 | * and should be short-lived. */ | |
b411b363 PR |
754 | void drbd_suspend_io(struct drbd_conf *mdev) |
755 | { | |
756 | set_bit(SUSPEND_IO, &mdev->flags); | |
2aebfabb | 757 | if (drbd_suspended(mdev)) |
265be2d0 | 758 | return; |
b411b363 PR |
759 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt)); |
760 | } | |
761 | ||
762 | void drbd_resume_io(struct drbd_conf *mdev) | |
763 | { | |
764 | clear_bit(SUSPEND_IO, &mdev->flags); | |
765 | wake_up(&mdev->misc_wait); | |
766 | } | |
767 | ||
768 | /** | |
769 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
770 | * @mdev: DRBD device. | |
771 | * | |
772 | * Returns 0 on success, negative return values indicate errors. | |
773 | * You should call drbd_md_sync() after calling this function. | |
774 | */ | |
24c4830c | 775 | enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local) |
b411b363 PR |
776 | { |
777 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
778 | sector_t la_size; | |
779 | sector_t size; | |
780 | char ppb[10]; | |
781 | ||
782 | int md_moved, la_size_changed; | |
783 | enum determine_dev_size rv = unchanged; | |
784 | ||
785 | /* race: | |
786 | * application request passes inc_ap_bio, | |
787 | * but then cannot get an AL-reference. | |
788 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
789 | * | |
790 | * to avoid that: | |
791 | * Suspend IO right here. | |
792 | * still lock the act_log to not trigger ASSERTs there. | |
793 | */ | |
794 | drbd_suspend_io(mdev); | |
795 | ||
796 | /* no wait necessary anymore, actually we could assert that */ | |
797 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); | |
798 | ||
799 | prev_first_sect = drbd_md_first_sector(mdev->ldev); | |
800 | prev_size = mdev->ldev->md.md_size_sect; | |
801 | la_size = mdev->ldev->md.la_size_sect; | |
802 | ||
803 | /* TODO: should only be some assert here, not (re)init... */ | |
804 | drbd_md_set_sector_offsets(mdev, mdev->ldev); | |
805 | ||
d845030f | 806 | size = drbd_new_dev_size(mdev, mdev->ldev, flags & DDSF_FORCED); |
b411b363 PR |
807 | |
808 | if (drbd_get_capacity(mdev->this_bdev) != size || | |
809 | drbd_bm_capacity(mdev) != size) { | |
810 | int err; | |
02d9a94b | 811 | err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
812 | if (unlikely(err)) { |
813 | /* currently there is only one error: ENOMEM! */ | |
814 | size = drbd_bm_capacity(mdev)>>1; | |
815 | if (size == 0) { | |
816 | dev_err(DEV, "OUT OF MEMORY! " | |
817 | "Could not allocate bitmap!\n"); | |
818 | } else { | |
819 | dev_err(DEV, "BM resizing failed. " | |
820 | "Leaving size unchanged at size = %lu KB\n", | |
821 | (unsigned long)size); | |
822 | } | |
823 | rv = dev_size_error; | |
824 | } | |
825 | /* racy, see comments above. */ | |
826 | drbd_set_my_capacity(mdev, size); | |
827 | mdev->ldev->md.la_size_sect = size; | |
828 | dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), | |
829 | (unsigned long long)size>>1); | |
830 | } | |
831 | if (rv == dev_size_error) | |
832 | goto out; | |
833 | ||
834 | la_size_changed = (la_size != mdev->ldev->md.la_size_sect); | |
835 | ||
836 | md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev) | |
837 | || prev_size != mdev->ldev->md.md_size_sect; | |
838 | ||
839 | if (la_size_changed || md_moved) { | |
24dccabb AG |
840 | int err; |
841 | ||
b411b363 PR |
842 | drbd_al_shrink(mdev); /* All extents inactive. */ |
843 | dev_info(DEV, "Writing the whole bitmap, %s\n", | |
844 | la_size_changed && md_moved ? "size changed and md moved" : | |
845 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 LE |
846 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
847 | err = drbd_bitmap_io(mdev, &drbd_bm_write, | |
848 | "size changed", BM_LOCKED_MASK); | |
24dccabb AG |
849 | if (err) { |
850 | rv = dev_size_error; | |
851 | goto out; | |
852 | } | |
b411b363 PR |
853 | drbd_md_mark_dirty(mdev); |
854 | } | |
855 | ||
856 | if (size > la_size) | |
857 | rv = grew; | |
858 | if (size < la_size) | |
859 | rv = shrunk; | |
860 | out: | |
861 | lc_unlock(mdev->act_log); | |
862 | wake_up(&mdev->al_wait); | |
863 | drbd_resume_io(mdev); | |
864 | ||
865 | return rv; | |
866 | } | |
867 | ||
868 | sector_t | |
a393db6f | 869 | drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, int assume_peer_has_space) |
b411b363 PR |
870 | { |
871 | sector_t p_size = mdev->p_size; /* partner's disk size. */ | |
872 | sector_t la_size = bdev->md.la_size_sect; /* last agreed size. */ | |
873 | sector_t m_size; /* my size */ | |
874 | sector_t u_size = bdev->dc.disk_size; /* size requested by user. */ | |
875 | sector_t size = 0; | |
876 | ||
877 | m_size = drbd_get_max_capacity(bdev); | |
878 | ||
a393db6f PR |
879 | if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) { |
880 | dev_warn(DEV, "Resize while not connected was forced by the user!\n"); | |
881 | p_size = m_size; | |
882 | } | |
883 | ||
b411b363 PR |
884 | if (p_size && m_size) { |
885 | size = min_t(sector_t, p_size, m_size); | |
886 | } else { | |
887 | if (la_size) { | |
888 | size = la_size; | |
889 | if (m_size && m_size < size) | |
890 | size = m_size; | |
891 | if (p_size && p_size < size) | |
892 | size = p_size; | |
893 | } else { | |
894 | if (m_size) | |
895 | size = m_size; | |
896 | if (p_size) | |
897 | size = p_size; | |
898 | } | |
899 | } | |
900 | ||
901 | if (size == 0) | |
902 | dev_err(DEV, "Both nodes diskless!\n"); | |
903 | ||
904 | if (u_size) { | |
905 | if (u_size > size) | |
906 | dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n", | |
907 | (unsigned long)u_size>>1, (unsigned long)size>>1); | |
908 | else | |
909 | size = u_size; | |
910 | } | |
911 | ||
912 | return size; | |
913 | } | |
914 | ||
915 | /** | |
916 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
917 | * @mdev: DRBD device. | |
918 | * | |
919 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
920 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
921 | * this function. | |
922 | */ | |
f399002e | 923 | static int drbd_check_al_size(struct drbd_conf *mdev, struct disk_conf *dc) |
b411b363 PR |
924 | { |
925 | struct lru_cache *n, *t; | |
926 | struct lc_element *e; | |
927 | unsigned int in_use; | |
928 | int i; | |
929 | ||
f399002e LE |
930 | if (!expect(dc->al_extents >= DRBD_AL_EXTENTS_MIN)) |
931 | dc->al_extents = DRBD_AL_EXTENTS_MIN; | |
b411b363 PR |
932 | |
933 | if (mdev->act_log && | |
f399002e | 934 | mdev->act_log->nr_elements == dc->al_extents) |
b411b363 PR |
935 | return 0; |
936 | ||
937 | in_use = 0; | |
938 | t = mdev->act_log; | |
7ad651b5 | 939 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 940 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
941 | |
942 | if (n == NULL) { | |
943 | dev_err(DEV, "Cannot allocate act_log lru!\n"); | |
944 | return -ENOMEM; | |
945 | } | |
946 | spin_lock_irq(&mdev->al_lock); | |
947 | if (t) { | |
948 | for (i = 0; i < t->nr_elements; i++) { | |
949 | e = lc_element_by_index(t, i); | |
950 | if (e->refcnt) | |
951 | dev_err(DEV, "refcnt(%d)==%d\n", | |
952 | e->lc_number, e->refcnt); | |
953 | in_use += e->refcnt; | |
954 | } | |
955 | } | |
956 | if (!in_use) | |
957 | mdev->act_log = n; | |
958 | spin_unlock_irq(&mdev->al_lock); | |
959 | if (in_use) { | |
960 | dev_err(DEV, "Activity log still in use!\n"); | |
961 | lc_destroy(n); | |
962 | return -EBUSY; | |
963 | } else { | |
964 | if (t) | |
965 | lc_destroy(t); | |
966 | } | |
967 | drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */ | |
968 | return 0; | |
969 | } | |
970 | ||
99432fcc | 971 | static void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size) |
b411b363 PR |
972 | { |
973 | struct request_queue * const q = mdev->rq_queue; | |
99432fcc PR |
974 | int max_hw_sectors = max_bio_size >> 9; |
975 | int max_segments = 0; | |
976 | ||
977 | if (get_ldev_if_state(mdev, D_ATTACHING)) { | |
978 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
979 | ||
980 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
981 | max_segments = mdev->ldev->dc.max_bio_bvecs; | |
982 | put_ldev(mdev); | |
983 | } | |
b411b363 | 984 | |
b411b363 | 985 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
986 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
987 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
988 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
989 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 990 | |
99432fcc PR |
991 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
992 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
993 | ||
994 | blk_queue_stack_limits(q, b); | |
995 | ||
996 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
997 | dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", | |
998 | q->backing_dev_info.ra_pages, | |
999 | b->backing_dev_info.ra_pages); | |
1000 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1001 | } | |
1002 | put_ldev(mdev); | |
1003 | } | |
1004 | } | |
1005 | ||
1006 | void drbd_reconsider_max_bio_size(struct drbd_conf *mdev) | |
1007 | { | |
1008 | int now, new, local, peer; | |
1009 | ||
1010 | now = queue_max_hw_sectors(mdev->rq_queue) << 9; | |
1011 | local = mdev->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1012 | peer = mdev->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1013 | |
99432fcc PR |
1014 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
1015 | local = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9; | |
1016 | mdev->local_max_bio_size = local; | |
1017 | put_ldev(mdev); | |
b411b363 | 1018 | } |
99432fcc PR |
1019 | |
1020 | /* We may ignore peer limits if the peer is modern enough. | |
1021 | Because new from 8.3.8 onwards the peer can use multiple | |
1022 | BIOs for a single peer_request */ | |
1023 | if (mdev->state.conn >= C_CONNECTED) { | |
31890f4a | 1024 | if (mdev->tconn->agreed_pro_version < 94) |
99432fcc | 1025 | peer = mdev->peer_max_bio_size; |
31890f4a | 1026 | else if (mdev->tconn->agreed_pro_version == 94) |
99432fcc PR |
1027 | peer = DRBD_MAX_SIZE_H80_PACKET; |
1028 | else /* drbd 8.3.8 onwards */ | |
1029 | peer = DRBD_MAX_BIO_SIZE; | |
1030 | } | |
1031 | ||
1032 | new = min_t(int, local, peer); | |
1033 | ||
1034 | if (mdev->state.role == R_PRIMARY && new < now) | |
1035 | dev_err(DEV, "ASSERT FAILED new < now; (%d < %d)\n", new, now); | |
1036 | ||
1037 | if (new != now) | |
1038 | dev_info(DEV, "max BIO size = %u\n", new); | |
1039 | ||
1040 | drbd_setup_queue_param(mdev, new); | |
b411b363 PR |
1041 | } |
1042 | ||
a18e9d1e | 1043 | /* Starts the worker thread */ |
0e29d163 | 1044 | static void conn_reconfig_start(struct drbd_tconn *tconn) |
b411b363 | 1045 | { |
0e29d163 PR |
1046 | drbd_thread_start(&tconn->worker); |
1047 | conn_flush_workqueue(tconn); | |
b411b363 PR |
1048 | } |
1049 | ||
a18e9d1e | 1050 | /* if still unconfigured, stops worker again. */ |
0e29d163 | 1051 | static void conn_reconfig_done(struct drbd_tconn *tconn) |
b411b363 | 1052 | { |
992d6e91 | 1053 | bool stop_threads; |
0e29d163 | 1054 | spin_lock_irq(&tconn->req_lock); |
992d6e91 | 1055 | stop_threads = conn_all_vols_unconf(tconn); |
0e29d163 | 1056 | spin_unlock_irq(&tconn->req_lock); |
992d6e91 LE |
1057 | if (stop_threads) { |
1058 | /* asender is implicitly stopped by receiver | |
1059 | * in drbd_disconnect() */ | |
1060 | drbd_thread_stop(&tconn->receiver); | |
1061 | drbd_thread_stop(&tconn->worker); | |
1062 | } | |
b411b363 PR |
1063 | } |
1064 | ||
0778286a PR |
1065 | /* Make sure IO is suspended before calling this function(). */ |
1066 | static void drbd_suspend_al(struct drbd_conf *mdev) | |
1067 | { | |
1068 | int s = 0; | |
1069 | ||
61610420 | 1070 | if (!lc_try_lock(mdev->act_log)) { |
0778286a PR |
1071 | dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n"); |
1072 | return; | |
1073 | } | |
1074 | ||
61610420 | 1075 | drbd_al_shrink(mdev); |
87eeee41 | 1076 | spin_lock_irq(&mdev->tconn->req_lock); |
0778286a PR |
1077 | if (mdev->state.conn < C_CONNECTED) |
1078 | s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags); | |
87eeee41 | 1079 | spin_unlock_irq(&mdev->tconn->req_lock); |
61610420 | 1080 | lc_unlock(mdev->act_log); |
0778286a PR |
1081 | |
1082 | if (s) | |
1083 | dev_info(DEV, "Suspended AL updates\n"); | |
1084 | } | |
1085 | ||
5979e361 LE |
1086 | |
1087 | static bool should_set_defaults(struct genl_info *info) | |
1088 | { | |
1089 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1090 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1091 | } | |
1092 | ||
f399002e LE |
1093 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1094 | { | |
1095 | enum drbd_ret_code retcode; | |
1096 | struct drbd_conf *mdev; | |
5ecc72c3 | 1097 | struct disk_conf *new_disk_conf; |
f399002e LE |
1098 | int err, fifo_size; |
1099 | int *rs_plan_s = NULL; | |
1100 | ||
1101 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1102 | if (!adm_ctx.reply_skb) | |
1103 | return retcode; | |
1104 | if (retcode != NO_ERROR) | |
1105 | goto out; | |
1106 | ||
1107 | mdev = adm_ctx.mdev; | |
1108 | ||
1109 | /* we also need a disk | |
1110 | * to change the options on */ | |
1111 | if (!get_ldev(mdev)) { | |
1112 | retcode = ERR_NO_DISK; | |
1113 | goto out; | |
1114 | } | |
1115 | ||
1116 | /* FIXME freeze IO, cluster wide. | |
1117 | * | |
1118 | * We should make sure no-one uses | |
1119 | * some half-updated struct when we | |
1120 | * assign it later. */ | |
1121 | ||
5ecc72c3 LE |
1122 | new_disk_conf = kmalloc(sizeof(*new_disk_conf), GFP_KERNEL); |
1123 | if (!new_disk_conf) { | |
f399002e LE |
1124 | retcode = ERR_NOMEM; |
1125 | goto fail; | |
1126 | } | |
1127 | ||
5ecc72c3 | 1128 | memcpy(new_disk_conf, &mdev->ldev->dc, sizeof(*new_disk_conf)); |
5979e361 | 1129 | if (should_set_defaults(info)) |
b966b5dd | 1130 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1131 | |
5ecc72c3 | 1132 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
f399002e LE |
1133 | if (err) { |
1134 | retcode = ERR_MANDATORY_TAG; | |
1135 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1136 | } | |
1137 | ||
5ecc72c3 LE |
1138 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1139 | new_disk_conf->resync_rate = 1; | |
f399002e LE |
1140 | |
1141 | /* clip to allowed range */ | |
5ecc72c3 LE |
1142 | if (!expect(new_disk_conf->al_extents >= DRBD_AL_EXTENTS_MIN)) |
1143 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1144 | if (!expect(new_disk_conf->al_extents <= DRBD_AL_EXTENTS_MAX)) | |
1145 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MAX; | |
f399002e LE |
1146 | |
1147 | /* most sanity checks done, try to assign the new sync-after | |
1148 | * dependency. need to hold the global lock in there, | |
1149 | * to avoid a race in the dependency loop check. */ | |
5ecc72c3 | 1150 | retcode = drbd_alter_sa(mdev, new_disk_conf->resync_after); |
f399002e LE |
1151 | if (retcode != NO_ERROR) |
1152 | goto fail; | |
1153 | ||
5ecc72c3 | 1154 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
f399002e LE |
1155 | if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) { |
1156 | rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL); | |
1157 | if (!rs_plan_s) { | |
1158 | dev_err(DEV, "kmalloc of fifo_buffer failed"); | |
1159 | retcode = ERR_NOMEM; | |
1160 | goto fail; | |
1161 | } | |
1162 | } | |
1163 | ||
1164 | if (fifo_size != mdev->rs_plan_s.size) { | |
1165 | kfree(mdev->rs_plan_s.values); | |
1166 | mdev->rs_plan_s.values = rs_plan_s; | |
1167 | mdev->rs_plan_s.size = fifo_size; | |
1168 | mdev->rs_planed = 0; | |
1169 | rs_plan_s = NULL; | |
1170 | } | |
1171 | ||
1172 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); | |
1173 | drbd_al_shrink(mdev); | |
5ecc72c3 | 1174 | err = drbd_check_al_size(mdev, new_disk_conf); |
f399002e LE |
1175 | lc_unlock(mdev->act_log); |
1176 | wake_up(&mdev->al_wait); | |
1177 | ||
1178 | if (err) { | |
1179 | retcode = ERR_NOMEM; | |
1180 | goto fail; | |
1181 | } | |
1182 | ||
1183 | /* FIXME | |
1184 | * To avoid someone looking at a half-updated struct, we probably | |
1185 | * should have a rw-semaphor on net_conf and disk_conf. | |
1186 | */ | |
5ecc72c3 | 1187 | mdev->ldev->dc = *new_disk_conf; |
f399002e LE |
1188 | |
1189 | drbd_md_sync(mdev); | |
1190 | ||
1191 | ||
1192 | if (mdev->state.conn >= C_CONNECTED) | |
1193 | drbd_send_sync_param(mdev); | |
1194 | ||
1195 | fail: | |
1196 | put_ldev(mdev); | |
5ecc72c3 | 1197 | kfree(new_disk_conf); |
f399002e LE |
1198 | kfree(rs_plan_s); |
1199 | out: | |
1200 | drbd_adm_finish(info, retcode); | |
1201 | return 0; | |
1202 | } | |
1203 | ||
3b98c0c2 | 1204 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1205 | { |
3b98c0c2 LE |
1206 | struct drbd_conf *mdev; |
1207 | int err; | |
116676ca | 1208 | enum drbd_ret_code retcode; |
b411b363 PR |
1209 | enum determine_dev_size dd; |
1210 | sector_t max_possible_sectors; | |
1211 | sector_t min_md_device_sectors; | |
1212 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
e525fd89 | 1213 | struct block_device *bdev; |
b411b363 PR |
1214 | struct lru_cache *resync_lru = NULL; |
1215 | union drbd_state ns, os; | |
f2024e7c | 1216 | enum drbd_state_rv rv; |
44ed167d | 1217 | struct net_conf *nc; |
b411b363 | 1218 | int cp_discovered = 0; |
b411b363 | 1219 | |
3b98c0c2 LE |
1220 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1221 | if (!adm_ctx.reply_skb) | |
1222 | return retcode; | |
1223 | if (retcode != NO_ERROR) | |
40cbf085 | 1224 | goto finish; |
3b98c0c2 LE |
1225 | |
1226 | mdev = adm_ctx.mdev; | |
0e29d163 | 1227 | conn_reconfig_start(mdev->tconn); |
b411b363 PR |
1228 | |
1229 | /* if you want to reconfigure, please tear down first */ | |
1230 | if (mdev->state.disk > D_DISKLESS) { | |
1231 | retcode = ERR_DISK_CONFIGURED; | |
1232 | goto fail; | |
1233 | } | |
82f59cc6 LE |
1234 | /* It may just now have detached because of IO error. Make sure |
1235 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1236 | * e.g. if someone calls attach from the on-io-error handler, | |
1237 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
1238 | wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); | |
b411b363 | 1239 | |
3b98c0c2 | 1240 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1241 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1242 | if (!nbc) { | |
1243 | retcode = ERR_NOMEM; | |
1244 | goto fail; | |
1245 | } | |
1246 | ||
b966b5dd | 1247 | set_disk_conf_defaults(&nbc->dc); |
f399002e LE |
1248 | |
1249 | err = disk_conf_from_attrs(&nbc->dc, info); | |
3b98c0c2 | 1250 | if (err) { |
b411b363 | 1251 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1252 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1253 | goto fail; |
1254 | } | |
1255 | ||
3b98c0c2 | 1256 | if ((int)nbc->dc.meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1257 | retcode = ERR_MD_IDX_INVALID; |
1258 | goto fail; | |
1259 | } | |
1260 | ||
44ed167d PR |
1261 | rcu_read_lock(); |
1262 | nc = rcu_dereference(mdev->tconn->net_conf); | |
1263 | if (nc) { | |
1264 | if (nbc->dc.fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { | |
1265 | rcu_read_unlock(); | |
47ff2d0a PR |
1266 | retcode = ERR_STONITH_AND_PROT_A; |
1267 | goto fail; | |
1268 | } | |
1269 | } | |
44ed167d | 1270 | rcu_read_unlock(); |
47ff2d0a | 1271 | |
d4d77629 TH |
1272 | bdev = blkdev_get_by_path(nbc->dc.backing_dev, |
1273 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev); | |
e525fd89 | 1274 | if (IS_ERR(bdev)) { |
b411b363 | 1275 | dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.backing_dev, |
e525fd89 | 1276 | PTR_ERR(bdev)); |
b411b363 PR |
1277 | retcode = ERR_OPEN_DISK; |
1278 | goto fail; | |
1279 | } | |
e525fd89 TH |
1280 | nbc->backing_bdev = bdev; |
1281 | ||
1282 | /* | |
1283 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1284 | * drbd sharing one meta device. TODO in that case, paranoia | |
1285 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1286 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1287 | * should check it for you already; but if you don't, or | |
1288 | * someone fooled it, we need to double check here) | |
1289 | */ | |
d4d77629 TH |
1290 | bdev = blkdev_get_by_path(nbc->dc.meta_dev, |
1291 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, | |
3b98c0c2 | 1292 | ((int)nbc->dc.meta_dev_idx < 0) ? |
d4d77629 | 1293 | (void *)mdev : (void *)drbd_m_holder); |
e525fd89 | 1294 | if (IS_ERR(bdev)) { |
b411b363 | 1295 | dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.meta_dev, |
e525fd89 | 1296 | PTR_ERR(bdev)); |
b411b363 PR |
1297 | retcode = ERR_OPEN_MD_DISK; |
1298 | goto fail; | |
1299 | } | |
e525fd89 | 1300 | nbc->md_bdev = bdev; |
b411b363 | 1301 | |
e525fd89 TH |
1302 | if ((nbc->backing_bdev == nbc->md_bdev) != |
1303 | (nbc->dc.meta_dev_idx == DRBD_MD_INDEX_INTERNAL || | |
1304 | nbc->dc.meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
1305 | retcode = ERR_MD_IDX_INVALID; | |
b411b363 PR |
1306 | goto fail; |
1307 | } | |
1308 | ||
1309 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1310 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1311 | offsetof(struct bm_extent, lce)); |
1312 | if (!resync_lru) { | |
1313 | retcode = ERR_NOMEM; | |
e525fd89 | 1314 | goto fail; |
b411b363 PR |
1315 | } |
1316 | ||
1317 | /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */ | |
1318 | drbd_md_set_sector_offsets(mdev, nbc); | |
1319 | ||
1320 | if (drbd_get_max_capacity(nbc) < nbc->dc.disk_size) { | |
1321 | dev_err(DEV, "max capacity %llu smaller than disk size %llu\n", | |
1322 | (unsigned long long) drbd_get_max_capacity(nbc), | |
1323 | (unsigned long long) nbc->dc.disk_size); | |
1324 | retcode = ERR_DISK_TO_SMALL; | |
e525fd89 | 1325 | goto fail; |
b411b363 PR |
1326 | } |
1327 | ||
3b98c0c2 | 1328 | if ((int)nbc->dc.meta_dev_idx < 0) { |
b411b363 PR |
1329 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1330 | /* at least one MB, otherwise it does not make sense */ | |
1331 | min_md_device_sectors = (2<<10); | |
1332 | } else { | |
1333 | max_possible_sectors = DRBD_MAX_SECTORS; | |
1334 | min_md_device_sectors = MD_RESERVED_SECT * (nbc->dc.meta_dev_idx + 1); | |
1335 | } | |
1336 | ||
b411b363 PR |
1337 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
1338 | retcode = ERR_MD_DISK_TO_SMALL; | |
1339 | dev_warn(DEV, "refusing attach: md-device too small, " | |
1340 | "at least %llu sectors needed for this meta-disk type\n", | |
1341 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1342 | goto fail; |
b411b363 PR |
1343 | } |
1344 | ||
1345 | /* Make sure the new disk is big enough | |
1346 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1347 | if (drbd_get_max_capacity(nbc) < | |
1348 | drbd_get_capacity(mdev->this_bdev)) { | |
1349 | retcode = ERR_DISK_TO_SMALL; | |
e525fd89 | 1350 | goto fail; |
b411b363 PR |
1351 | } |
1352 | ||
1353 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1354 | ||
1352994b LE |
1355 | if (nbc->known_size > max_possible_sectors) { |
1356 | dev_warn(DEV, "==> truncating very big lower level device " | |
1357 | "to currently maximum possible %llu sectors <==\n", | |
1358 | (unsigned long long) max_possible_sectors); | |
3b98c0c2 | 1359 | if ((int)nbc->dc.meta_dev_idx >= 0) |
1352994b LE |
1360 | dev_warn(DEV, "==>> using internal or flexible " |
1361 | "meta data may help <<==\n"); | |
1362 | } | |
1363 | ||
b411b363 PR |
1364 | drbd_suspend_io(mdev); |
1365 | /* also wait for the last barrier ack. */ | |
2aebfabb | 1366 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || drbd_suspended(mdev)); |
b411b363 | 1367 | /* and for any other previously queued work */ |
a21e9298 | 1368 | drbd_flush_workqueue(mdev); |
b411b363 | 1369 | |
f2024e7c AG |
1370 | rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE); |
1371 | retcode = rv; /* FIXME: Type mismatch. */ | |
b411b363 | 1372 | drbd_resume_io(mdev); |
f2024e7c | 1373 | if (rv < SS_SUCCESS) |
e525fd89 | 1374 | goto fail; |
b411b363 PR |
1375 | |
1376 | if (!get_ldev_if_state(mdev, D_ATTACHING)) | |
1377 | goto force_diskless; | |
1378 | ||
1379 | drbd_md_set_sector_offsets(mdev, nbc); | |
1380 | ||
1381 | if (!mdev->bitmap) { | |
1382 | if (drbd_bm_init(mdev)) { | |
1383 | retcode = ERR_NOMEM; | |
1384 | goto force_diskless_dec; | |
1385 | } | |
1386 | } | |
1387 | ||
1388 | retcode = drbd_md_read(mdev, nbc); | |
1389 | if (retcode != NO_ERROR) | |
1390 | goto force_diskless_dec; | |
1391 | ||
1392 | if (mdev->state.conn < C_CONNECTED && | |
1393 | mdev->state.role == R_PRIMARY && | |
1394 | (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
1395 | dev_err(DEV, "Can only attach to data with current UUID=%016llX\n", | |
1396 | (unsigned long long)mdev->ed_uuid); | |
1397 | retcode = ERR_DATA_NOT_CURRENT; | |
1398 | goto force_diskless_dec; | |
1399 | } | |
1400 | ||
1401 | /* Since we are diskless, fix the activity log first... */ | |
f399002e | 1402 | if (drbd_check_al_size(mdev, &nbc->dc)) { |
b411b363 PR |
1403 | retcode = ERR_NOMEM; |
1404 | goto force_diskless_dec; | |
1405 | } | |
1406 | ||
1407 | /* Prevent shrinking of consistent devices ! */ | |
1408 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
a393db6f | 1409 | drbd_new_dev_size(mdev, nbc, 0) < nbc->md.la_size_sect) { |
b411b363 PR |
1410 | dev_warn(DEV, "refusing to truncate a consistent device\n"); |
1411 | retcode = ERR_DISK_TO_SMALL; | |
1412 | goto force_diskless_dec; | |
1413 | } | |
1414 | ||
1415 | if (!drbd_al_read_log(mdev, nbc)) { | |
1416 | retcode = ERR_IO_MD_DISK; | |
1417 | goto force_diskless_dec; | |
1418 | } | |
1419 | ||
b411b363 PR |
1420 | /* Reset the "barriers don't work" bits here, then force meta data to |
1421 | * be written, to ensure we determine if barriers are supported. */ | |
1422 | if (nbc->dc.no_md_flush) | |
a8a4e51e | 1423 | set_bit(MD_NO_FUA, &mdev->flags); |
b411b363 | 1424 | else |
a8a4e51e | 1425 | clear_bit(MD_NO_FUA, &mdev->flags); |
b411b363 PR |
1426 | |
1427 | /* Point of no return reached. | |
1428 | * Devices and memory are no longer released by error cleanup below. | |
1429 | * now mdev takes over responsibility, and the state engine should | |
1430 | * clean it up somewhere. */ | |
1431 | D_ASSERT(mdev->ldev == NULL); | |
1432 | mdev->ldev = nbc; | |
1433 | mdev->resync = resync_lru; | |
1434 | nbc = NULL; | |
1435 | resync_lru = NULL; | |
1436 | ||
2451fc3b PR |
1437 | mdev->write_ordering = WO_bdev_flush; |
1438 | drbd_bump_write_ordering(mdev, WO_bdev_flush); | |
b411b363 PR |
1439 | |
1440 | if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY)) | |
1441 | set_bit(CRASHED_PRIMARY, &mdev->flags); | |
1442 | else | |
1443 | clear_bit(CRASHED_PRIMARY, &mdev->flags); | |
1444 | ||
894c6a94 | 1445 | if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && |
da9fbc27 | 1446 | !(mdev->state.role == R_PRIMARY && mdev->tconn->susp_nod)) { |
b411b363 PR |
1447 | set_bit(CRASHED_PRIMARY, &mdev->flags); |
1448 | cp_discovered = 1; | |
1449 | } | |
1450 | ||
1451 | mdev->send_cnt = 0; | |
1452 | mdev->recv_cnt = 0; | |
1453 | mdev->read_cnt = 0; | |
1454 | mdev->writ_cnt = 0; | |
1455 | ||
99432fcc | 1456 | drbd_reconsider_max_bio_size(mdev); |
b411b363 PR |
1457 | |
1458 | /* If I am currently not R_PRIMARY, | |
1459 | * but meta data primary indicator is set, | |
1460 | * I just now recover from a hard crash, | |
1461 | * and have been R_PRIMARY before that crash. | |
1462 | * | |
1463 | * Now, if I had no connection before that crash | |
1464 | * (have been degraded R_PRIMARY), chances are that | |
1465 | * I won't find my peer now either. | |
1466 | * | |
1467 | * In that case, and _only_ in that case, | |
1468 | * we use the degr-wfc-timeout instead of the default, | |
1469 | * so we can automatically recover from a crash of a | |
1470 | * degraded but active "cluster" after a certain timeout. | |
1471 | */ | |
1472 | clear_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1473 | if (mdev->state.role != R_PRIMARY && | |
1474 | drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && | |
1475 | !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND)) | |
1476 | set_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1477 | ||
24c4830c | 1478 | dd = drbd_determine_dev_size(mdev, 0); |
b411b363 PR |
1479 | if (dd == dev_size_error) { |
1480 | retcode = ERR_NOMEM_BITMAP; | |
1481 | goto force_diskless_dec; | |
1482 | } else if (dd == grew) | |
1483 | set_bit(RESYNC_AFTER_NEG, &mdev->flags); | |
1484 | ||
1485 | if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) { | |
1486 | dev_info(DEV, "Assuming that all blocks are out of sync " | |
1487 | "(aka FullSync)\n"); | |
20ceb2b2 LE |
1488 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, |
1489 | "set_n_write from attaching", BM_LOCKED_MASK)) { | |
b411b363 PR |
1490 | retcode = ERR_IO_MD_DISK; |
1491 | goto force_diskless_dec; | |
1492 | } | |
1493 | } else { | |
20ceb2b2 | 1494 | if (drbd_bitmap_io(mdev, &drbd_bm_read, |
22ab6a30 | 1495 | "read from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1496 | retcode = ERR_IO_MD_DISK; |
1497 | goto force_diskless_dec; | |
1498 | } | |
1499 | } | |
1500 | ||
1501 | if (cp_discovered) { | |
1502 | drbd_al_apply_to_bm(mdev); | |
20ceb2b2 LE |
1503 | if (drbd_bitmap_io(mdev, &drbd_bm_write, |
1504 | "crashed primary apply AL", BM_LOCKED_MASK)) { | |
19f843aa LE |
1505 | retcode = ERR_IO_MD_DISK; |
1506 | goto force_diskless_dec; | |
1507 | } | |
b411b363 PR |
1508 | } |
1509 | ||
0778286a PR |
1510 | if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev)) |
1511 | drbd_suspend_al(mdev); /* IO is still suspended here... */ | |
1512 | ||
87eeee41 | 1513 | spin_lock_irq(&mdev->tconn->req_lock); |
78bae59b PR |
1514 | os = drbd_read_state(mdev); |
1515 | ns = os; | |
b411b363 PR |
1516 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1517 | otherwise investigate MDF_WasUpToDate... | |
1518 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1519 | otherwise into D_CONSISTENT state. | |
1520 | */ | |
1521 | if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) { | |
1522 | if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE)) | |
1523 | ns.disk = D_CONSISTENT; | |
1524 | else | |
1525 | ns.disk = D_OUTDATED; | |
1526 | } else { | |
1527 | ns.disk = D_INCONSISTENT; | |
1528 | } | |
1529 | ||
1530 | if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED)) | |
1531 | ns.pdsk = D_OUTDATED; | |
1532 | ||
1533 | if ( ns.disk == D_CONSISTENT && | |
1534 | (ns.pdsk == D_OUTDATED || mdev->ldev->dc.fencing == FP_DONT_CARE)) | |
1535 | ns.disk = D_UP_TO_DATE; | |
1536 | ||
1537 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1538 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1539 | this point, because drbd_request_state() modifies these | |
1540 | flags. */ | |
1541 | ||
1542 | /* In case we are C_CONNECTED postpone any decision on the new disk | |
1543 | state after the negotiation phase. */ | |
1544 | if (mdev->state.conn == C_CONNECTED) { | |
1545 | mdev->new_state_tmp.i = ns.i; | |
1546 | ns.i = os.i; | |
1547 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1548 | |
1549 | /* We expect to receive up-to-date UUIDs soon. | |
1550 | To avoid a race in receive_state, free p_uuid while | |
1551 | holding req_lock. I.e. atomic with the state change */ | |
1552 | kfree(mdev->p_uuid); | |
1553 | mdev->p_uuid = NULL; | |
b411b363 PR |
1554 | } |
1555 | ||
1556 | rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL); | |
87eeee41 | 1557 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
1558 | |
1559 | if (rv < SS_SUCCESS) | |
1560 | goto force_diskless_dec; | |
1561 | ||
1562 | if (mdev->state.role == R_PRIMARY) | |
1563 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
1564 | else | |
1565 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
1566 | ||
1567 | drbd_md_mark_dirty(mdev); | |
1568 | drbd_md_sync(mdev); | |
1569 | ||
1570 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
1571 | put_ldev(mdev); | |
0e29d163 | 1572 | conn_reconfig_done(mdev->tconn); |
3b98c0c2 | 1573 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1574 | return 0; |
1575 | ||
1576 | force_diskless_dec: | |
1577 | put_ldev(mdev); | |
1578 | force_diskless: | |
82f59cc6 | 1579 | drbd_force_state(mdev, NS(disk, D_FAILED)); |
b411b363 | 1580 | drbd_md_sync(mdev); |
b411b363 | 1581 | fail: |
40cbf085 | 1582 | conn_reconfig_done(mdev->tconn); |
b411b363 | 1583 | if (nbc) { |
e525fd89 TH |
1584 | if (nbc->backing_bdev) |
1585 | blkdev_put(nbc->backing_bdev, | |
1586 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1587 | if (nbc->md_bdev) | |
1588 | blkdev_put(nbc->md_bdev, | |
1589 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1590 | kfree(nbc); |
1591 | } | |
1592 | lc_destroy(resync_lru); | |
1593 | ||
40cbf085 | 1594 | finish: |
3b98c0c2 | 1595 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1596 | return 0; |
1597 | } | |
1598 | ||
85f75dd7 LE |
1599 | static int adm_detach(struct drbd_conf *mdev) |
1600 | { | |
19f83c76 | 1601 | enum drbd_state_rv retcode; |
009ba89d | 1602 | int ret; |
85f75dd7 | 1603 | drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */ |
009ba89d LE |
1604 | retcode = drbd_request_state(mdev, NS(disk, D_FAILED)); |
1605 | /* D_FAILED will transition to DISKLESS. */ | |
1606 | ret = wait_event_interruptible(mdev->misc_wait, | |
1607 | mdev->state.disk != D_FAILED); | |
85f75dd7 | 1608 | drbd_resume_io(mdev); |
009ba89d LE |
1609 | if ((int)retcode == (int)SS_IS_DISKLESS) |
1610 | retcode = SS_NOTHING_TO_DO; | |
1611 | if (ret) | |
1612 | retcode = ERR_INTR; | |
85f75dd7 LE |
1613 | return retcode; |
1614 | } | |
1615 | ||
82f59cc6 LE |
1616 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1617 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1618 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1619 | * internal references as well. | |
1620 | * Only then we have finally detached. */ | |
3b98c0c2 | 1621 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1622 | { |
9a0d9d03 | 1623 | enum drbd_ret_code retcode; |
3b98c0c2 LE |
1624 | |
1625 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1626 | if (!adm_ctx.reply_skb) | |
1627 | return retcode; | |
1628 | if (retcode != NO_ERROR) | |
1629 | goto out; | |
1630 | ||
85f75dd7 | 1631 | retcode = adm_detach(adm_ctx.mdev); |
3b98c0c2 LE |
1632 | out: |
1633 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1634 | return 0; |
1635 | } | |
1636 | ||
f399002e LE |
1637 | static bool conn_resync_running(struct drbd_tconn *tconn) |
1638 | { | |
1639 | struct drbd_conf *mdev; | |
695d08fa | 1640 | bool rv = false; |
f399002e LE |
1641 | int vnr; |
1642 | ||
695d08fa | 1643 | rcu_read_lock(); |
f399002e LE |
1644 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1645 | if (mdev->state.conn == C_SYNC_SOURCE || | |
1646 | mdev->state.conn == C_SYNC_TARGET || | |
1647 | mdev->state.conn == C_PAUSED_SYNC_S || | |
695d08fa PR |
1648 | mdev->state.conn == C_PAUSED_SYNC_T) { |
1649 | rv = true; | |
1650 | break; | |
1651 | } | |
f399002e | 1652 | } |
695d08fa PR |
1653 | rcu_read_unlock(); |
1654 | ||
1655 | return rv; | |
f399002e LE |
1656 | } |
1657 | ||
1658 | static bool conn_ov_running(struct drbd_tconn *tconn) | |
1659 | { | |
1660 | struct drbd_conf *mdev; | |
695d08fa | 1661 | bool rv = false; |
f399002e LE |
1662 | int vnr; |
1663 | ||
695d08fa | 1664 | rcu_read_lock(); |
f399002e LE |
1665 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1666 | if (mdev->state.conn == C_VERIFY_S || | |
695d08fa PR |
1667 | mdev->state.conn == C_VERIFY_T) { |
1668 | rv = true; | |
1669 | break; | |
1670 | } | |
f399002e | 1671 | } |
695d08fa PR |
1672 | rcu_read_unlock(); |
1673 | ||
1674 | return rv; | |
f399002e LE |
1675 | } |
1676 | ||
cd64397c | 1677 | static enum drbd_ret_code |
44ed167d | 1678 | _check_net_options(struct drbd_tconn *tconn, struct net_conf *old_conf, struct net_conf *new_conf) |
cd64397c PR |
1679 | { |
1680 | struct drbd_conf *mdev; | |
1681 | int i; | |
1682 | ||
44ed167d | 1683 | if (old_conf && tconn->agreed_pro_version < 100 && |
b032b6fa | 1684 | tconn->cstate == C_WF_REPORT_PARAMS && |
44ed167d | 1685 | new_conf->wire_protocol != old_conf->wire_protocol) |
b032b6fa PR |
1686 | return ERR_NEED_APV_100; |
1687 | ||
cd64397c PR |
1688 | if (new_conf->two_primaries && |
1689 | (new_conf->wire_protocol != DRBD_PROT_C)) | |
1690 | return ERR_NOT_PROTO_C; | |
1691 | ||
cd64397c PR |
1692 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
1693 | if (get_ldev(mdev)) { | |
1694 | enum drbd_fencing_p fp = mdev->ldev->dc.fencing; | |
1695 | put_ldev(mdev); | |
44ed167d | 1696 | if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1697 | return ERR_STONITH_AND_PROT_A; |
cd64397c | 1698 | } |
44ed167d | 1699 | if (mdev->state.role == R_PRIMARY && new_conf->want_lose) |
cd64397c | 1700 | return ERR_DISCARD; |
cd64397c | 1701 | } |
cd64397c PR |
1702 | |
1703 | if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A) | |
1704 | return ERR_CONG_NOT_PROTO_A; | |
1705 | ||
1706 | return NO_ERROR; | |
1707 | } | |
1708 | ||
44ed167d PR |
1709 | static enum drbd_ret_code |
1710 | check_net_options(struct drbd_tconn *tconn, struct net_conf *new_conf) | |
1711 | { | |
1712 | static enum drbd_ret_code rv; | |
1713 | struct drbd_conf *mdev; | |
1714 | int i; | |
1715 | ||
1716 | rcu_read_lock(); | |
1717 | rv = _check_net_options(tconn, rcu_dereference(tconn->net_conf), new_conf); | |
1718 | rcu_read_unlock(); | |
1719 | ||
1720 | /* tconn->volumes protected by genl_lock() here */ | |
1721 | idr_for_each_entry(&tconn->volumes, mdev, i) { | |
1722 | if (!mdev->bitmap) { | |
1723 | if(drbd_bm_init(mdev)) | |
1724 | return ERR_NOMEM; | |
1725 | } | |
1726 | } | |
1727 | ||
1728 | return rv; | |
1729 | } | |
1730 | ||
0fd0ea06 PR |
1731 | struct crypto { |
1732 | struct crypto_hash *verify_tfm; | |
1733 | struct crypto_hash *csums_tfm; | |
1734 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 1735 | struct crypto_hash *integrity_tfm; |
0fd0ea06 PR |
1736 | void *int_dig_in; |
1737 | void *int_dig_vv; | |
1738 | }; | |
1739 | ||
1740 | static int | |
4b6ad6d4 | 1741 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
1742 | { |
1743 | if (!tfm_name[0]) | |
1744 | return NO_ERROR; | |
1745 | ||
1746 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); | |
1747 | if (IS_ERR(*tfm)) { | |
1748 | *tfm = NULL; | |
1749 | return err_alg; | |
1750 | } | |
1751 | ||
0fd0ea06 PR |
1752 | return NO_ERROR; |
1753 | } | |
1754 | ||
1755 | static enum drbd_ret_code | |
1756 | alloc_crypto(struct crypto *crypto, struct net_conf *new_conf) | |
1757 | { | |
1758 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
1759 | enum drbd_ret_code rv; | |
1760 | int hash_size; | |
1761 | ||
4b6ad6d4 AG |
1762 | rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg, |
1763 | ERR_CSUMS_ALG); | |
0fd0ea06 PR |
1764 | if (rv != NO_ERROR) |
1765 | return rv; | |
4b6ad6d4 AG |
1766 | rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg, |
1767 | ERR_VERIFY_ALG); | |
0fd0ea06 PR |
1768 | if (rv != NO_ERROR) |
1769 | return rv; | |
4b6ad6d4 AG |
1770 | rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg, |
1771 | ERR_INTEGRITY_ALG); | |
0fd0ea06 PR |
1772 | if (rv != NO_ERROR) |
1773 | return rv; | |
0fd0ea06 PR |
1774 | if (new_conf->cram_hmac_alg[0] != 0) { |
1775 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", | |
1776 | new_conf->cram_hmac_alg); | |
1777 | ||
4b6ad6d4 AG |
1778 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
1779 | ERR_AUTH_ALG); | |
0fd0ea06 | 1780 | } |
8d412fc6 AG |
1781 | if (crypto->integrity_tfm) { |
1782 | hash_size = crypto_hash_digestsize(crypto->integrity_tfm); | |
0fd0ea06 PR |
1783 | crypto->int_dig_in = kmalloc(hash_size, GFP_KERNEL); |
1784 | if (!crypto->int_dig_in) | |
1785 | return ERR_NOMEM; | |
1786 | crypto->int_dig_vv = kmalloc(hash_size, GFP_KERNEL); | |
1787 | if (!crypto->int_dig_vv) | |
1788 | return ERR_NOMEM; | |
1789 | } | |
1790 | ||
1791 | return rv; | |
1792 | } | |
1793 | ||
1794 | static void free_crypto(struct crypto *crypto) | |
1795 | { | |
1796 | kfree(crypto->int_dig_in); | |
1797 | kfree(crypto->int_dig_vv); | |
1798 | crypto_free_hash(crypto->cram_hmac_tfm); | |
8d412fc6 | 1799 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
1800 | crypto_free_hash(crypto->csums_tfm); |
1801 | crypto_free_hash(crypto->verify_tfm); | |
1802 | } | |
1803 | ||
f399002e LE |
1804 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
1805 | { | |
1806 | enum drbd_ret_code retcode; | |
1807 | struct drbd_tconn *tconn; | |
44ed167d | 1808 | struct net_conf *old_conf, *new_conf = NULL; |
f399002e LE |
1809 | int err; |
1810 | int ovr; /* online verify running */ | |
1811 | int rsr; /* re-sync running */ | |
0fd0ea06 | 1812 | struct crypto crypto = { }; |
88104ca4 | 1813 | bool change_integrity_alg; |
f399002e LE |
1814 | |
1815 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); | |
1816 | if (!adm_ctx.reply_skb) | |
1817 | return retcode; | |
1818 | if (retcode != NO_ERROR) | |
1819 | goto out; | |
1820 | ||
1821 | tconn = adm_ctx.tconn; | |
1822 | ||
1823 | new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); | |
1824 | if (!new_conf) { | |
1825 | retcode = ERR_NOMEM; | |
1826 | goto out; | |
1827 | } | |
1828 | ||
44ed167d PR |
1829 | conn_reconfig_start(tconn); |
1830 | ||
88104ca4 | 1831 | mutex_lock(&tconn->data.mutex); |
91fd4dad PR |
1832 | mutex_lock(&tconn->net_conf_update); |
1833 | old_conf = tconn->net_conf; | |
44ed167d PR |
1834 | |
1835 | if (!old_conf) { | |
f399002e LE |
1836 | drbd_msg_put_info("net conf missing, try connect"); |
1837 | retcode = ERR_INVALID_REQUEST; | |
91fd4dad | 1838 | goto fail; |
f399002e LE |
1839 | } |
1840 | ||
44ed167d | 1841 | *new_conf = *old_conf; |
5979e361 | 1842 | if (should_set_defaults(info)) |
b966b5dd | 1843 | set_net_conf_defaults(new_conf); |
f399002e | 1844 | |
f399002e LE |
1845 | err = net_conf_from_attrs_for_change(new_conf, info); |
1846 | if (err) { | |
1847 | retcode = ERR_MANDATORY_TAG; | |
1848 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1849 | goto fail; | |
1850 | } | |
1851 | ||
cd64397c PR |
1852 | retcode = check_net_options(tconn, new_conf); |
1853 | if (retcode != NO_ERROR) | |
1854 | goto fail; | |
1855 | ||
f399002e LE |
1856 | /* re-sync running */ |
1857 | rsr = conn_resync_running(tconn); | |
0fd0ea06 | 1858 | if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) { |
f399002e | 1859 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 1860 | goto fail; |
f399002e LE |
1861 | } |
1862 | ||
f399002e LE |
1863 | /* online verify running */ |
1864 | ovr = conn_ov_running(tconn); | |
0fd0ea06 PR |
1865 | if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) { |
1866 | retcode = ERR_VERIFY_RUNNING; | |
1867 | goto fail; | |
f399002e LE |
1868 | } |
1869 | ||
88104ca4 AG |
1870 | change_integrity_alg = strcmp(old_conf->integrity_alg, |
1871 | new_conf->integrity_alg); | |
1872 | ||
0fd0ea06 PR |
1873 | retcode = alloc_crypto(&crypto, new_conf); |
1874 | if (retcode != NO_ERROR) | |
1875 | goto fail; | |
f399002e | 1876 | |
44ed167d | 1877 | rcu_assign_pointer(tconn->net_conf, new_conf); |
f399002e LE |
1878 | |
1879 | if (!rsr) { | |
1880 | crypto_free_hash(tconn->csums_tfm); | |
0fd0ea06 PR |
1881 | tconn->csums_tfm = crypto.csums_tfm; |
1882 | crypto.csums_tfm = NULL; | |
f399002e LE |
1883 | } |
1884 | if (!ovr) { | |
1885 | crypto_free_hash(tconn->verify_tfm); | |
0fd0ea06 PR |
1886 | tconn->verify_tfm = crypto.verify_tfm; |
1887 | crypto.verify_tfm = NULL; | |
f399002e LE |
1888 | } |
1889 | ||
0fd0ea06 PR |
1890 | kfree(tconn->int_dig_in); |
1891 | tconn->int_dig_in = crypto.int_dig_in; | |
1892 | kfree(tconn->int_dig_vv); | |
1893 | tconn->int_dig_vv = crypto.int_dig_vv; | |
8d412fc6 AG |
1894 | crypto_free_hash(tconn->integrity_tfm); |
1895 | tconn->integrity_tfm = crypto.integrity_tfm; | |
88104ca4 AG |
1896 | if (change_integrity_alg) { |
1897 | /* Do this without trying to take tconn->data.mutex again. */ | |
1898 | if (__drbd_send_protocol(tconn)) | |
1899 | goto fail; | |
1900 | } | |
0fd0ea06 PR |
1901 | |
1902 | /* FIXME Changing cram_hmac while the connection is established is useless */ | |
1903 | crypto_free_hash(tconn->cram_hmac_tfm); | |
1904 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
1905 | ||
91fd4dad | 1906 | mutex_unlock(&tconn->net_conf_update); |
88104ca4 | 1907 | mutex_unlock(&tconn->data.mutex); |
91fd4dad PR |
1908 | synchronize_rcu(); |
1909 | kfree(old_conf); | |
1910 | ||
f399002e LE |
1911 | if (tconn->cstate >= C_WF_REPORT_PARAMS) |
1912 | drbd_send_sync_param(minor_to_mdev(conn_lowest_minor(tconn))); | |
1913 | ||
91fd4dad PR |
1914 | goto done; |
1915 | ||
f399002e | 1916 | fail: |
91fd4dad | 1917 | mutex_unlock(&tconn->net_conf_update); |
88104ca4 | 1918 | mutex_unlock(&tconn->data.mutex); |
0fd0ea06 | 1919 | free_crypto(&crypto); |
f399002e | 1920 | kfree(new_conf); |
91fd4dad | 1921 | done: |
f399002e LE |
1922 | conn_reconfig_done(tconn); |
1923 | out: | |
1924 | drbd_adm_finish(info, retcode); | |
1925 | return 0; | |
1926 | } | |
1927 | ||
3b98c0c2 | 1928 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1929 | { |
3b98c0c2 | 1930 | struct drbd_conf *mdev; |
44ed167d | 1931 | struct net_conf *old_conf, *new_conf = NULL; |
0fd0ea06 | 1932 | struct crypto crypto = { }; |
80883197 | 1933 | struct drbd_tconn *oconn; |
3b98c0c2 | 1934 | struct drbd_tconn *tconn; |
b411b363 | 1935 | struct sockaddr *new_my_addr, *new_peer_addr, *taken_addr; |
3b98c0c2 LE |
1936 | enum drbd_ret_code retcode; |
1937 | int i; | |
1938 | int err; | |
b411b363 | 1939 | |
3b98c0c2 LE |
1940 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
1941 | if (!adm_ctx.reply_skb) | |
1942 | return retcode; | |
1943 | if (retcode != NO_ERROR) | |
1944 | goto out; | |
1945 | ||
1946 | tconn = adm_ctx.tconn; | |
80883197 | 1947 | conn_reconfig_start(tconn); |
b411b363 | 1948 | |
80883197 | 1949 | if (tconn->cstate > C_STANDALONE) { |
b411b363 PR |
1950 | retcode = ERR_NET_CONFIGURED; |
1951 | goto fail; | |
1952 | } | |
1953 | ||
1954 | /* allocation not in the IO path, cqueue thread context */ | |
5979e361 | 1955 | new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL); |
b411b363 PR |
1956 | if (!new_conf) { |
1957 | retcode = ERR_NOMEM; | |
1958 | goto fail; | |
1959 | } | |
1960 | ||
b966b5dd | 1961 | set_net_conf_defaults(new_conf); |
f399002e LE |
1962 | |
1963 | err = net_conf_from_attrs(new_conf, info); | |
3b98c0c2 | 1964 | if (err) { |
b411b363 | 1965 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1966 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1967 | goto fail; |
1968 | } | |
1969 | ||
cd64397c PR |
1970 | retcode = check_net_options(tconn, new_conf); |
1971 | if (retcode != NO_ERROR) | |
422028b1 | 1972 | goto fail; |
422028b1 | 1973 | |
b411b363 PR |
1974 | retcode = NO_ERROR; |
1975 | ||
1976 | new_my_addr = (struct sockaddr *)&new_conf->my_addr; | |
1977 | new_peer_addr = (struct sockaddr *)&new_conf->peer_addr; | |
543cc10b | 1978 | |
ef356262 | 1979 | /* No need to take drbd_cfg_rwsem here. All reconfiguration is |
543cc10b LE |
1980 | * strictly serialized on genl_lock(). We are protected against |
1981 | * concurrent reconfiguration/addition/deletion */ | |
80883197 | 1982 | list_for_each_entry(oconn, &drbd_tconns, all_tconn) { |
44ed167d | 1983 | struct net_conf *nc; |
80883197 | 1984 | if (oconn == tconn) |
b411b363 | 1985 | continue; |
44ed167d PR |
1986 | |
1987 | rcu_read_lock(); | |
1988 | nc = rcu_dereference(oconn->net_conf); | |
1989 | if (nc) { | |
1990 | taken_addr = (struct sockaddr *)&nc->my_addr; | |
1991 | if (new_conf->my_addr_len == nc->my_addr_len && | |
b411b363 PR |
1992 | !memcmp(new_my_addr, taken_addr, new_conf->my_addr_len)) |
1993 | retcode = ERR_LOCAL_ADDR; | |
1994 | ||
44ed167d PR |
1995 | taken_addr = (struct sockaddr *)&nc->peer_addr; |
1996 | if (new_conf->peer_addr_len == nc->peer_addr_len && | |
b411b363 PR |
1997 | !memcmp(new_peer_addr, taken_addr, new_conf->peer_addr_len)) |
1998 | retcode = ERR_PEER_ADDR; | |
b411b363 | 1999 | } |
44ed167d PR |
2000 | rcu_read_unlock(); |
2001 | if (retcode != NO_ERROR) | |
2002 | goto fail; | |
b411b363 PR |
2003 | } |
2004 | ||
0fd0ea06 PR |
2005 | retcode = alloc_crypto(&crypto, new_conf); |
2006 | if (retcode != NO_ERROR) | |
2007 | goto fail; | |
b411b363 | 2008 | |
b411b363 PR |
2009 | ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
2010 | ||
80883197 | 2011 | conn_flush_workqueue(tconn); |
91fd4dad PR |
2012 | |
2013 | mutex_lock(&tconn->net_conf_update); | |
2014 | old_conf = tconn->net_conf; | |
2015 | if (old_conf) { | |
b411b363 | 2016 | retcode = ERR_NET_CONFIGURED; |
91fd4dad | 2017 | mutex_unlock(&tconn->net_conf_update); |
b411b363 PR |
2018 | goto fail; |
2019 | } | |
44ed167d | 2020 | rcu_assign_pointer(tconn->net_conf, new_conf); |
b411b363 | 2021 | |
91fd4dad | 2022 | conn_free_crypto(tconn); |
0fd0ea06 PR |
2023 | tconn->int_dig_in = crypto.int_dig_in; |
2024 | tconn->int_dig_vv = crypto.int_dig_vv; | |
2025 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
8d412fc6 | 2026 | tconn->integrity_tfm = crypto.integrity_tfm; |
0fd0ea06 PR |
2027 | tconn->csums_tfm = crypto.csums_tfm; |
2028 | tconn->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2029 | |
91fd4dad PR |
2030 | mutex_unlock(&tconn->net_conf_update); |
2031 | ||
695d08fa | 2032 | rcu_read_lock(); |
80883197 PR |
2033 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
2034 | mdev->send_cnt = 0; | |
2035 | mdev->recv_cnt = 0; | |
80883197 | 2036 | } |
695d08fa | 2037 | rcu_read_unlock(); |
5ee743e9 LE |
2038 | |
2039 | retcode = conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE); | |
2040 | ||
80883197 | 2041 | conn_reconfig_done(tconn); |
3b98c0c2 | 2042 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2043 | return 0; |
2044 | ||
2045 | fail: | |
0fd0ea06 | 2046 | free_crypto(&crypto); |
b411b363 PR |
2047 | kfree(new_conf); |
2048 | ||
80883197 | 2049 | conn_reconfig_done(tconn); |
3b98c0c2 LE |
2050 | out: |
2051 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2052 | return 0; |
2053 | } | |
2054 | ||
85f75dd7 LE |
2055 | static enum drbd_state_rv conn_try_disconnect(struct drbd_tconn *tconn, bool force) |
2056 | { | |
2057 | enum drbd_state_rv rv; | |
85f75dd7 | 2058 | |
f3dfa40a LE |
2059 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2060 | force ? CS_HARD : 0); | |
85f75dd7 LE |
2061 | |
2062 | switch (rv) { | |
2063 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2064 | break; |
85f75dd7 LE |
2065 | case SS_ALREADY_STANDALONE: |
2066 | return SS_SUCCESS; | |
2067 | case SS_PRIMARY_NOP: | |
2068 | /* Our state checking code wants to see the peer outdated. */ | |
2069 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
f3dfa40a | 2070 | pdsk, D_OUTDATED), CS_VERBOSE); |
85f75dd7 LE |
2071 | break; |
2072 | case SS_CW_FAILED_BY_PEER: | |
2073 | /* The peer probably wants to see us outdated. */ | |
2074 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
2075 | disk, D_OUTDATED), 0); | |
2076 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
f3dfa40a LE |
2077 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2078 | CS_HARD); | |
85f75dd7 LE |
2079 | } |
2080 | break; | |
2081 | default:; | |
2082 | /* no special handling necessary */ | |
2083 | } | |
2084 | ||
f3dfa40a LE |
2085 | if (rv >= SS_SUCCESS) { |
2086 | enum drbd_state_rv rv2; | |
2087 | /* No one else can reconfigure the network while I am here. | |
2088 | * The state handling only uses drbd_thread_stop_nowait(), | |
2089 | * we want to really wait here until the receiver is no more. | |
2090 | */ | |
2091 | drbd_thread_stop(&adm_ctx.tconn->receiver); | |
2092 | ||
2093 | /* Race breaker. This additional state change request may be | |
2094 | * necessary, if this was a forced disconnect during a receiver | |
2095 | * restart. We may have "killed" the receiver thread just | |
2096 | * after drbdd_init() returned. Typically, we should be | |
2097 | * C_STANDALONE already, now, and this becomes a no-op. | |
2098 | */ | |
2099 | rv2 = conn_request_state(tconn, NS(conn, C_STANDALONE), | |
2100 | CS_VERBOSE | CS_HARD); | |
2101 | if (rv2 < SS_SUCCESS) | |
2102 | conn_err(tconn, | |
2103 | "unexpected rv2=%d in conn_try_disconnect()\n", | |
2104 | rv2); | |
2105 | } | |
85f75dd7 LE |
2106 | return rv; |
2107 | } | |
2108 | ||
3b98c0c2 | 2109 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2110 | { |
3b98c0c2 LE |
2111 | struct disconnect_parms parms; |
2112 | struct drbd_tconn *tconn; | |
85f75dd7 | 2113 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2114 | enum drbd_ret_code retcode; |
2115 | int err; | |
2561b9c1 | 2116 | |
3b98c0c2 LE |
2117 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
2118 | if (!adm_ctx.reply_skb) | |
2119 | return retcode; | |
2120 | if (retcode != NO_ERROR) | |
2561b9c1 | 2121 | goto fail; |
3b98c0c2 LE |
2122 | |
2123 | tconn = adm_ctx.tconn; | |
2124 | memset(&parms, 0, sizeof(parms)); | |
2125 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2126 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2127 | if (err) { |
2128 | retcode = ERR_MANDATORY_TAG; | |
2129 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2130 | goto fail; | |
2131 | } | |
2561b9c1 PR |
2132 | } |
2133 | ||
85f75dd7 LE |
2134 | rv = conn_try_disconnect(tconn, parms.force_disconnect); |
2135 | if (rv < SS_SUCCESS) | |
f3dfa40a LE |
2136 | retcode = rv; /* FIXME: Type mismatch. */ |
2137 | else | |
2138 | retcode = NO_ERROR; | |
b411b363 | 2139 | fail: |
3b98c0c2 | 2140 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2141 | return 0; |
2142 | } | |
2143 | ||
2144 | void resync_after_online_grow(struct drbd_conf *mdev) | |
2145 | { | |
2146 | int iass; /* I am sync source */ | |
2147 | ||
2148 | dev_info(DEV, "Resync of new storage after online grow\n"); | |
2149 | if (mdev->state.role != mdev->state.peer) | |
2150 | iass = (mdev->state.role == R_PRIMARY); | |
2151 | else | |
25703f83 | 2152 | iass = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags); |
b411b363 PR |
2153 | |
2154 | if (iass) | |
2155 | drbd_start_resync(mdev, C_SYNC_SOURCE); | |
2156 | else | |
2157 | _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); | |
2158 | } | |
2159 | ||
3b98c0c2 | 2160 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2161 | { |
3b98c0c2 LE |
2162 | struct resize_parms rs; |
2163 | struct drbd_conf *mdev; | |
2164 | enum drbd_ret_code retcode; | |
b411b363 | 2165 | enum determine_dev_size dd; |
6495d2c6 | 2166 | enum dds_flags ddsf; |
3b98c0c2 | 2167 | int err; |
b411b363 | 2168 | |
3b98c0c2 LE |
2169 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2170 | if (!adm_ctx.reply_skb) | |
2171 | return retcode; | |
2172 | if (retcode != NO_ERROR) | |
b411b363 | 2173 | goto fail; |
3b98c0c2 LE |
2174 | |
2175 | memset(&rs, 0, sizeof(struct resize_parms)); | |
2176 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { | |
f399002e | 2177 | err = resize_parms_from_attrs(&rs, info); |
3b98c0c2 LE |
2178 | if (err) { |
2179 | retcode = ERR_MANDATORY_TAG; | |
2180 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2181 | goto fail; | |
2182 | } | |
b411b363 PR |
2183 | } |
2184 | ||
3b98c0c2 | 2185 | mdev = adm_ctx.mdev; |
b411b363 PR |
2186 | if (mdev->state.conn > C_CONNECTED) { |
2187 | retcode = ERR_RESIZE_RESYNC; | |
2188 | goto fail; | |
2189 | } | |
2190 | ||
2191 | if (mdev->state.role == R_SECONDARY && | |
2192 | mdev->state.peer == R_SECONDARY) { | |
2193 | retcode = ERR_NO_PRIMARY; | |
2194 | goto fail; | |
2195 | } | |
2196 | ||
2197 | if (!get_ldev(mdev)) { | |
2198 | retcode = ERR_NO_DISK; | |
2199 | goto fail; | |
2200 | } | |
2201 | ||
31890f4a | 2202 | if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) { |
6495d2c6 PR |
2203 | retcode = ERR_NEED_APV_93; |
2204 | goto fail; | |
2205 | } | |
2206 | ||
087c2492 | 2207 | if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) |
b411b363 | 2208 | mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); |
b411b363 PR |
2209 | |
2210 | mdev->ldev->dc.disk_size = (sector_t)rs.resize_size; | |
6495d2c6 | 2211 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
24c4830c | 2212 | dd = drbd_determine_dev_size(mdev, ddsf); |
b411b363 PR |
2213 | drbd_md_sync(mdev); |
2214 | put_ldev(mdev); | |
2215 | if (dd == dev_size_error) { | |
2216 | retcode = ERR_NOMEM_BITMAP; | |
2217 | goto fail; | |
2218 | } | |
2219 | ||
087c2492 | 2220 | if (mdev->state.conn == C_CONNECTED) { |
b411b363 PR |
2221 | if (dd == grew) |
2222 | set_bit(RESIZE_PENDING, &mdev->flags); | |
2223 | ||
2224 | drbd_send_uuids(mdev); | |
6495d2c6 | 2225 | drbd_send_sizes(mdev, 1, ddsf); |
b411b363 PR |
2226 | } |
2227 | ||
2228 | fail: | |
3b98c0c2 | 2229 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2230 | return 0; |
2231 | } | |
2232 | ||
b966b5dd AG |
2233 | void drbd_set_res_opts_defaults(struct res_opts *r) |
2234 | { | |
2235 | return set_res_opts_defaults(r); | |
2236 | } | |
2237 | ||
f399002e | 2238 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2239 | { |
3b98c0c2 | 2240 | enum drbd_ret_code retcode; |
b411b363 | 2241 | cpumask_var_t new_cpu_mask; |
f399002e | 2242 | struct drbd_tconn *tconn; |
b57a1e27 | 2243 | struct res_opts res_opts; |
f399002e | 2244 | int err; |
b411b363 | 2245 | |
f399002e | 2246 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
3b98c0c2 LE |
2247 | if (!adm_ctx.reply_skb) |
2248 | return retcode; | |
2249 | if (retcode != NO_ERROR) | |
2250 | goto fail; | |
f399002e | 2251 | tconn = adm_ctx.tconn; |
3b98c0c2 | 2252 | |
b411b363 PR |
2253 | if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL)) { |
2254 | retcode = ERR_NOMEM; | |
3b98c0c2 | 2255 | drbd_msg_put_info("unable to allocate cpumask"); |
b411b363 PR |
2256 | goto fail; |
2257 | } | |
2258 | ||
b57a1e27 | 2259 | res_opts = tconn->res_opts; |
5979e361 | 2260 | if (should_set_defaults(info)) |
b966b5dd | 2261 | set_res_opts_defaults(&res_opts); |
b411b363 | 2262 | |
b57a1e27 | 2263 | err = res_opts_from_attrs(&res_opts, info); |
3b98c0c2 | 2264 | if (err) { |
b411b363 | 2265 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2266 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2267 | goto fail; |
2268 | } | |
2269 | ||
b411b363 | 2270 | /* silently ignore cpu mask on UP kernel */ |
b57a1e27 LE |
2271 | if (nr_cpu_ids > 1 && res_opts.cpu_mask[0] != 0) { |
2272 | err = __bitmap_parse(res_opts.cpu_mask, 32, 0, | |
b411b363 PR |
2273 | cpumask_bits(new_cpu_mask), nr_cpu_ids); |
2274 | if (err) { | |
f399002e | 2275 | conn_warn(tconn, "__bitmap_parse() failed with %d\n", err); |
b411b363 PR |
2276 | retcode = ERR_CPU_MASK_PARSE; |
2277 | goto fail; | |
2278 | } | |
2279 | } | |
2280 | ||
b411b363 | 2281 | |
b57a1e27 | 2282 | tconn->res_opts = res_opts; |
b411b363 | 2283 | |
f399002e LE |
2284 | if (!cpumask_equal(tconn->cpu_mask, new_cpu_mask)) { |
2285 | cpumask_copy(tconn->cpu_mask, new_cpu_mask); | |
2286 | drbd_calc_cpu_mask(tconn); | |
2287 | tconn->receiver.reset_cpu_mask = 1; | |
2288 | tconn->asender.reset_cpu_mask = 1; | |
2289 | tconn->worker.reset_cpu_mask = 1; | |
b411b363 PR |
2290 | } |
2291 | ||
b411b363 PR |
2292 | fail: |
2293 | free_cpumask_var(new_cpu_mask); | |
3b98c0c2 LE |
2294 | |
2295 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2296 | return 0; |
2297 | } | |
2298 | ||
3b98c0c2 | 2299 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2300 | { |
3b98c0c2 LE |
2301 | struct drbd_conf *mdev; |
2302 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2303 | ||
2304 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2305 | if (!adm_ctx.reply_skb) | |
2306 | return retcode; | |
2307 | if (retcode != NO_ERROR) | |
2308 | goto out; | |
2309 | ||
2310 | mdev = adm_ctx.mdev; | |
b411b363 | 2311 | |
194bfb32 LE |
2312 | /* If there is still bitmap IO pending, probably because of a previous |
2313 | * resync just being finished, wait for it before requesting a new resync. */ | |
2314 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); | |
2315 | ||
b411b363 PR |
2316 | retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED); |
2317 | ||
2318 | if (retcode < SS_SUCCESS && retcode != SS_NEED_CONNECTION) | |
2319 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T)); | |
2320 | ||
2321 | while (retcode == SS_NEED_CONNECTION) { | |
87eeee41 | 2322 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2323 | if (mdev->state.conn < C_CONNECTED) |
2324 | retcode = _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_VERBOSE, NULL); | |
87eeee41 | 2325 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2326 | |
2327 | if (retcode != SS_NEED_CONNECTION) | |
2328 | break; | |
2329 | ||
2330 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T)); | |
2331 | } | |
2332 | ||
3b98c0c2 LE |
2333 | out: |
2334 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2335 | return 0; |
2336 | } | |
2337 | ||
0778286a PR |
2338 | static int drbd_bmio_set_susp_al(struct drbd_conf *mdev) |
2339 | { | |
2340 | int rv; | |
2341 | ||
2342 | rv = drbd_bmio_set_n_write(mdev); | |
2343 | drbd_suspend_al(mdev); | |
2344 | return rv; | |
2345 | } | |
2346 | ||
3b98c0c2 LE |
2347 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2348 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2349 | { |
3b98c0c2 | 2350 | enum drbd_ret_code retcode; |
194bfb32 | 2351 | |
3b98c0c2 LE |
2352 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2353 | if (!adm_ctx.reply_skb) | |
2354 | return retcode; | |
2355 | if (retcode != NO_ERROR) | |
2356 | goto out; | |
b411b363 | 2357 | |
3b98c0c2 LE |
2358 | retcode = drbd_request_state(adm_ctx.mdev, mask, val); |
2359 | out: | |
2360 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2361 | return 0; |
2362 | } | |
2363 | ||
3b98c0c2 | 2364 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2365 | { |
3b98c0c2 LE |
2366 | return drbd_adm_simple_request_state(skb, info, NS(conn, C_STARTING_SYNC_S)); |
2367 | } | |
b411b363 | 2368 | |
3b98c0c2 LE |
2369 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
2370 | { | |
2371 | enum drbd_ret_code retcode; | |
2372 | ||
2373 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2374 | if (!adm_ctx.reply_skb) | |
2375 | return retcode; | |
2376 | if (retcode != NO_ERROR) | |
2377 | goto out; | |
b411b363 | 2378 | |
3b98c0c2 LE |
2379 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
2380 | retcode = ERR_PAUSE_IS_SET; | |
2381 | out: | |
2382 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2383 | return 0; |
2384 | } | |
2385 | ||
3b98c0c2 | 2386 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2387 | { |
da9fbc27 | 2388 | union drbd_dev_state s; |
3b98c0c2 LE |
2389 | enum drbd_ret_code retcode; |
2390 | ||
2391 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2392 | if (!adm_ctx.reply_skb) | |
2393 | return retcode; | |
2394 | if (retcode != NO_ERROR) | |
2395 | goto out; | |
b411b363 | 2396 | |
3b98c0c2 LE |
2397 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2398 | s = adm_ctx.mdev->state; | |
cd88d030 PR |
2399 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2400 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2401 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2402 | } else { | |
2403 | retcode = ERR_PAUSE_IS_CLEAR; | |
2404 | } | |
2405 | } | |
b411b363 | 2406 | |
3b98c0c2 LE |
2407 | out: |
2408 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2409 | return 0; |
2410 | } | |
2411 | ||
3b98c0c2 | 2412 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2413 | { |
3b98c0c2 | 2414 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2415 | } |
2416 | ||
3b98c0c2 | 2417 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2418 | { |
3b98c0c2 LE |
2419 | struct drbd_conf *mdev; |
2420 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2421 | ||
2422 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2423 | if (!adm_ctx.reply_skb) | |
2424 | return retcode; | |
2425 | if (retcode != NO_ERROR) | |
2426 | goto out; | |
2427 | ||
2428 | mdev = adm_ctx.mdev; | |
43a5182c PR |
2429 | if (test_bit(NEW_CUR_UUID, &mdev->flags)) { |
2430 | drbd_uuid_new_current(mdev); | |
2431 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
43a5182c | 2432 | } |
265be2d0 | 2433 | drbd_suspend_io(mdev); |
3b98c0c2 LE |
2434 | retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); |
2435 | if (retcode == SS_SUCCESS) { | |
265be2d0 | 2436 | if (mdev->state.conn < C_CONNECTED) |
2f5cdd0b | 2437 | tl_clear(mdev->tconn); |
265be2d0 | 2438 | if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED) |
2f5cdd0b | 2439 | tl_restart(mdev->tconn, FAIL_FROZEN_DISK_IO); |
265be2d0 PR |
2440 | } |
2441 | drbd_resume_io(mdev); | |
2442 | ||
3b98c0c2 LE |
2443 | out: |
2444 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2445 | return 0; |
2446 | } | |
2447 | ||
3b98c0c2 | 2448 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2449 | { |
3b98c0c2 | 2450 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2451 | } |
2452 | ||
543cc10b LE |
2453 | int nla_put_drbd_cfg_context(struct sk_buff *skb, const char *conn_name, unsigned vnr) |
2454 | { | |
2455 | struct nlattr *nla; | |
2456 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2457 | if (!nla) | |
2458 | goto nla_put_failure; | |
2459 | if (vnr != VOLUME_UNSPECIFIED) | |
2460 | NLA_PUT_U32(skb, T_ctx_volume, vnr); | |
2461 | NLA_PUT_STRING(skb, T_ctx_conn_name, conn_name); | |
2462 | nla_nest_end(skb, nla); | |
2463 | return 0; | |
2464 | ||
2465 | nla_put_failure: | |
2466 | if (nla) | |
2467 | nla_nest_cancel(skb, nla); | |
2468 | return -EMSGSIZE; | |
2469 | } | |
2470 | ||
3b98c0c2 LE |
2471 | int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev, |
2472 | const struct sib_info *sib) | |
b411b363 | 2473 | { |
3b98c0c2 | 2474 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
44ed167d | 2475 | struct net_conf *nc; |
3b98c0c2 LE |
2476 | struct nlattr *nla; |
2477 | int got_ldev; | |
3b98c0c2 LE |
2478 | int err = 0; |
2479 | int exclude_sensitive; | |
2480 | ||
2481 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2482 | * to. So we better exclude_sensitive information. | |
2483 | * | |
2484 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2485 | * in the context of the requesting user process. Exclude sensitive | |
2486 | * information, unless current has superuser. | |
2487 | * | |
2488 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2489 | * relies on the current implementation of netlink_dump(), which | |
2490 | * executes the dump callback successively from netlink_recvmsg(), | |
2491 | * always in the context of the receiving process */ | |
2492 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2493 | ||
2494 | got_ldev = get_ldev(mdev); | |
3b98c0c2 LE |
2495 | |
2496 | /* We need to add connection name and volume number information still. | |
2497 | * Minor number is in drbd_genlmsghdr. */ | |
543cc10b | 2498 | if (nla_put_drbd_cfg_context(skb, mdev->tconn->name, mdev->vnr)) |
3b98c0c2 | 2499 | goto nla_put_failure; |
3b98c0c2 | 2500 | |
f399002e LE |
2501 | if (res_opts_to_skb(skb, &mdev->tconn->res_opts, exclude_sensitive)) |
2502 | goto nla_put_failure; | |
2503 | ||
3b98c0c2 LE |
2504 | if (got_ldev) |
2505 | if (disk_conf_to_skb(skb, &mdev->ldev->dc, exclude_sensitive)) | |
2506 | goto nla_put_failure; | |
44ed167d PR |
2507 | |
2508 | rcu_read_lock(); | |
2509 | nc = rcu_dereference(mdev->tconn->net_conf); | |
2510 | if (nc) | |
2511 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2512 | rcu_read_unlock(); | |
2513 | if (err) | |
2514 | goto nla_put_failure; | |
3b98c0c2 | 2515 | |
3b98c0c2 LE |
2516 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2517 | if (!nla) | |
2518 | goto nla_put_failure; | |
2519 | NLA_PUT_U32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY); | |
2520 | NLA_PUT_U32(skb, T_current_state, mdev->state.i); | |
2521 | NLA_PUT_U64(skb, T_ed_uuid, mdev->ed_uuid); | |
2522 | NLA_PUT_U64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev)); | |
2523 | ||
2524 | if (got_ldev) { | |
2525 | NLA_PUT_U32(skb, T_disk_flags, mdev->ldev->md.flags); | |
2526 | NLA_PUT(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid); | |
2527 | NLA_PUT_U64(skb, T_bits_total, drbd_bm_bits(mdev)); | |
2528 | NLA_PUT_U64(skb, T_bits_oos, drbd_bm_total_weight(mdev)); | |
2529 | if (C_SYNC_SOURCE <= mdev->state.conn && | |
2530 | C_PAUSED_SYNC_T >= mdev->state.conn) { | |
2531 | NLA_PUT_U64(skb, T_bits_rs_total, mdev->rs_total); | |
2532 | NLA_PUT_U64(skb, T_bits_rs_failed, mdev->rs_failed); | |
2533 | } | |
b411b363 PR |
2534 | } |
2535 | ||
3b98c0c2 LE |
2536 | if (sib) { |
2537 | switch(sib->sib_reason) { | |
2538 | case SIB_SYNC_PROGRESS: | |
2539 | case SIB_GET_STATUS_REPLY: | |
2540 | break; | |
2541 | case SIB_STATE_CHANGE: | |
2542 | NLA_PUT_U32(skb, T_prev_state, sib->os.i); | |
2543 | NLA_PUT_U32(skb, T_new_state, sib->ns.i); | |
2544 | break; | |
2545 | case SIB_HELPER_POST: | |
2546 | NLA_PUT_U32(skb, | |
2547 | T_helper_exit_code, sib->helper_exit_code); | |
2548 | /* fall through */ | |
2549 | case SIB_HELPER_PRE: | |
2550 | NLA_PUT_STRING(skb, T_helper, sib->helper_name); | |
2551 | break; | |
2552 | } | |
b411b363 | 2553 | } |
3b98c0c2 | 2554 | nla_nest_end(skb, nla); |
b411b363 | 2555 | |
3b98c0c2 LE |
2556 | if (0) |
2557 | nla_put_failure: | |
2558 | err = -EMSGSIZE; | |
2559 | if (got_ldev) | |
2560 | put_ldev(mdev); | |
3b98c0c2 | 2561 | return err; |
b411b363 PR |
2562 | } |
2563 | ||
3b98c0c2 | 2564 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2565 | { |
3b98c0c2 LE |
2566 | enum drbd_ret_code retcode; |
2567 | int err; | |
b411b363 | 2568 | |
3b98c0c2 LE |
2569 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2570 | if (!adm_ctx.reply_skb) | |
2571 | return retcode; | |
2572 | if (retcode != NO_ERROR) | |
2573 | goto out; | |
b411b363 | 2574 | |
3b98c0c2 LE |
2575 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL); |
2576 | if (err) { | |
2577 | nlmsg_free(adm_ctx.reply_skb); | |
2578 | return err; | |
b411b363 | 2579 | } |
3b98c0c2 LE |
2580 | out: |
2581 | drbd_adm_finish(info, retcode); | |
2582 | return 0; | |
b411b363 PR |
2583 | } |
2584 | ||
71932efc | 2585 | int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2586 | { |
3b98c0c2 LE |
2587 | struct drbd_conf *mdev; |
2588 | struct drbd_genlmsghdr *dh; | |
543cc10b LE |
2589 | struct drbd_tconn *pos = (struct drbd_tconn*)cb->args[0]; |
2590 | struct drbd_tconn *tconn = NULL; | |
2591 | struct drbd_tconn *tmp; | |
2592 | unsigned volume = cb->args[1]; | |
2593 | ||
2594 | /* Open coded, deferred, iteration: | |
2595 | * list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) { | |
2596 | * idr_for_each_entry(&tconn->volumes, mdev, i) { | |
2597 | * ... | |
2598 | * } | |
2599 | * } | |
2600 | * where tconn is cb->args[0]; | |
2601 | * and i is cb->args[1]; | |
2602 | * | |
71932efc LE |
2603 | * cb->args[2] indicates if we shall loop over all resources, |
2604 | * or just dump all volumes of a single resource. | |
2605 | * | |
3b98c0c2 LE |
2606 | * This may miss entries inserted after this dump started, |
2607 | * or entries deleted before they are reached. | |
543cc10b LE |
2608 | * |
2609 | * We need to make sure the mdev won't disappear while | |
2610 | * we are looking at it, and revalidate our iterators | |
2611 | * on each iteration. | |
2612 | */ | |
3b98c0c2 | 2613 | |
9dc9fbb3 | 2614 | /* synchronize with conn_create()/conn_destroy() */ |
ef356262 | 2615 | down_read(&drbd_cfg_rwsem); |
543cc10b LE |
2616 | /* revalidate iterator position */ |
2617 | list_for_each_entry(tmp, &drbd_tconns, all_tconn) { | |
2618 | if (pos == NULL) { | |
2619 | /* first iteration */ | |
2620 | pos = tmp; | |
2621 | tconn = pos; | |
2622 | break; | |
2623 | } | |
2624 | if (tmp == pos) { | |
2625 | tconn = pos; | |
2626 | break; | |
2627 | } | |
2628 | } | |
2629 | if (tconn) { | |
71932efc | 2630 | next_tconn: |
543cc10b LE |
2631 | mdev = idr_get_next(&tconn->volumes, &volume); |
2632 | if (!mdev) { | |
2633 | /* No more volumes to dump on this tconn. | |
2634 | * Advance tconn iterator. */ | |
2635 | pos = list_entry(tconn->all_tconn.next, | |
2636 | struct drbd_tconn, all_tconn); | |
71932efc | 2637 | /* Did we dump any volume on this tconn yet? */ |
543cc10b | 2638 | if (volume != 0) { |
71932efc LE |
2639 | /* If we reached the end of the list, |
2640 | * or only a single resource dump was requested, | |
2641 | * we are done. */ | |
2642 | if (&pos->all_tconn == &drbd_tconns || cb->args[2]) | |
2643 | goto out; | |
543cc10b | 2644 | volume = 0; |
71932efc | 2645 | tconn = pos; |
543cc10b LE |
2646 | goto next_tconn; |
2647 | } | |
2648 | } | |
2649 | ||
3b98c0c2 LE |
2650 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, |
2651 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
2652 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
2653 | if (!dh) | |
543cc10b LE |
2654 | goto out; |
2655 | ||
2656 | if (!mdev) { | |
2657 | /* this is a tconn without a single volume */ | |
2658 | dh->minor = -1U; | |
2659 | dh->ret_code = NO_ERROR; | |
2660 | if (nla_put_drbd_cfg_context(skb, tconn->name, VOLUME_UNSPECIFIED)) | |
2661 | genlmsg_cancel(skb, dh); | |
2662 | else | |
2663 | genlmsg_end(skb, dh); | |
2664 | goto out; | |
2665 | } | |
3b98c0c2 | 2666 | |
543cc10b LE |
2667 | D_ASSERT(mdev->vnr == volume); |
2668 | D_ASSERT(mdev->tconn == tconn); | |
3b98c0c2 | 2669 | |
543cc10b | 2670 | dh->minor = mdev_to_minor(mdev); |
3b98c0c2 LE |
2671 | dh->ret_code = NO_ERROR; |
2672 | ||
2673 | if (nla_put_status_info(skb, mdev, NULL)) { | |
2674 | genlmsg_cancel(skb, dh); | |
543cc10b | 2675 | goto out; |
3b98c0c2 LE |
2676 | } |
2677 | genlmsg_end(skb, dh); | |
2678 | } | |
b411b363 | 2679 | |
543cc10b | 2680 | out: |
ef356262 | 2681 | up_read(&drbd_cfg_rwsem); |
543cc10b LE |
2682 | /* where to start the next iteration */ |
2683 | cb->args[0] = (long)pos; | |
2684 | cb->args[1] = (pos == tconn) ? volume + 1 : 0; | |
b411b363 | 2685 | |
543cc10b LE |
2686 | /* No more tconns/volumes/minors found results in an empty skb. |
2687 | * Which will terminate the dump. */ | |
3b98c0c2 | 2688 | return skb->len; |
b411b363 PR |
2689 | } |
2690 | ||
71932efc LE |
2691 | /* |
2692 | * Request status of all resources, or of all volumes within a single resource. | |
2693 | * | |
2694 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
2695 | * Which means we cannot use the family->attrbuf or other such members, because | |
2696 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
2697 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
2698 | * | |
2699 | * Once things are setup properly, we call into get_one_status(). | |
2700 | */ | |
2701 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) | |
2702 | { | |
2703 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; | |
2704 | struct nlattr *nla; | |
2705 | const char *conn_name; | |
2706 | struct drbd_tconn *tconn; | |
2707 | ||
2708 | /* Is this a followup call? */ | |
2709 | if (cb->args[0]) { | |
2710 | /* ... of a single resource dump, | |
2711 | * and the resource iterator has been advanced already? */ | |
2712 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
2713 | return 0; /* DONE. */ | |
2714 | goto dump; | |
2715 | } | |
2716 | ||
2717 | /* First call (from netlink_dump_start). We need to figure out | |
2718 | * which resource(s) the user wants us to dump. */ | |
2719 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
2720 | nlmsg_attrlen(cb->nlh, hdrlen), | |
2721 | DRBD_NLA_CFG_CONTEXT); | |
2722 | ||
2723 | /* No explicit context given. Dump all. */ | |
2724 | if (!nla) | |
2725 | goto dump; | |
2726 | nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name)); | |
2727 | /* context given, but no name present? */ | |
2728 | if (!nla) | |
2729 | return -EINVAL; | |
2730 | conn_name = nla_data(nla); | |
0ace9dfa PR |
2731 | tconn = conn_get_by_name(conn_name); |
2732 | ||
71932efc LE |
2733 | if (!tconn) |
2734 | return -ENODEV; | |
2735 | ||
0ace9dfa PR |
2736 | kref_put(&tconn->kref, &conn_destroy); /* get_one_status() (re)validates tconn by itself */ |
2737 | ||
71932efc LE |
2738 | /* prime iterators, and set "filter" mode mark: |
2739 | * only dump this tconn. */ | |
2740 | cb->args[0] = (long)tconn; | |
2741 | /* cb->args[1] = 0; passed in this way. */ | |
2742 | cb->args[2] = (long)tconn; | |
2743 | ||
2744 | dump: | |
2745 | return get_one_status(skb, cb); | |
2746 | } | |
2747 | ||
3b98c0c2 | 2748 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2749 | { |
3b98c0c2 LE |
2750 | enum drbd_ret_code retcode; |
2751 | struct timeout_parms tp; | |
2752 | int err; | |
b411b363 | 2753 | |
3b98c0c2 LE |
2754 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2755 | if (!adm_ctx.reply_skb) | |
2756 | return retcode; | |
2757 | if (retcode != NO_ERROR) | |
2758 | goto out; | |
b411b363 | 2759 | |
3b98c0c2 LE |
2760 | tp.timeout_type = |
2761 | adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : | |
2762 | test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED : | |
2763 | UT_DEFAULT; | |
b411b363 | 2764 | |
3b98c0c2 LE |
2765 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
2766 | if (err) { | |
2767 | nlmsg_free(adm_ctx.reply_skb); | |
2768 | return err; | |
2769 | } | |
2770 | out: | |
2771 | drbd_adm_finish(info, retcode); | |
2772 | return 0; | |
b411b363 PR |
2773 | } |
2774 | ||
3b98c0c2 | 2775 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2776 | { |
3b98c0c2 LE |
2777 | struct drbd_conf *mdev; |
2778 | enum drbd_ret_code retcode; | |
b411b363 | 2779 | |
3b98c0c2 LE |
2780 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2781 | if (!adm_ctx.reply_skb) | |
2782 | return retcode; | |
2783 | if (retcode != NO_ERROR) | |
2784 | goto out; | |
873b0d5f | 2785 | |
3b98c0c2 LE |
2786 | mdev = adm_ctx.mdev; |
2787 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { | |
2788 | /* resume from last known position, if possible */ | |
2789 | struct start_ov_parms parms = | |
2790 | { .ov_start_sector = mdev->ov_start_sector }; | |
f399002e | 2791 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2792 | if (err) { |
2793 | retcode = ERR_MANDATORY_TAG; | |
2794 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2795 | goto out; | |
2796 | } | |
2797 | /* w_make_ov_request expects position to be aligned */ | |
2798 | mdev->ov_start_sector = parms.ov_start_sector & ~BM_SECT_PER_BIT; | |
2799 | } | |
873b0d5f LE |
2800 | /* If there is still bitmap IO pending, e.g. previous resync or verify |
2801 | * just being finished, wait for it before requesting a new resync. */ | |
2802 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); | |
3b98c0c2 LE |
2803 | retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S)); |
2804 | out: | |
2805 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2806 | return 0; |
2807 | } | |
2808 | ||
2809 | ||
3b98c0c2 | 2810 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2811 | { |
3b98c0c2 LE |
2812 | struct drbd_conf *mdev; |
2813 | enum drbd_ret_code retcode; | |
b411b363 PR |
2814 | int skip_initial_sync = 0; |
2815 | int err; | |
3b98c0c2 | 2816 | struct new_c_uuid_parms args; |
b411b363 | 2817 | |
3b98c0c2 LE |
2818 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2819 | if (!adm_ctx.reply_skb) | |
2820 | return retcode; | |
2821 | if (retcode != NO_ERROR) | |
2822 | goto out_nolock; | |
b411b363 | 2823 | |
3b98c0c2 LE |
2824 | mdev = adm_ctx.mdev; |
2825 | memset(&args, 0, sizeof(args)); | |
2826 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 2827 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
2828 | if (err) { |
2829 | retcode = ERR_MANDATORY_TAG; | |
2830 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2831 | goto out_nolock; | |
2832 | } | |
b411b363 PR |
2833 | } |
2834 | ||
8410da8f | 2835 | mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 PR |
2836 | |
2837 | if (!get_ldev(mdev)) { | |
2838 | retcode = ERR_NO_DISK; | |
2839 | goto out; | |
2840 | } | |
2841 | ||
2842 | /* this is "skip initial sync", assume to be clean */ | |
31890f4a | 2843 | if (mdev->state.conn == C_CONNECTED && mdev->tconn->agreed_pro_version >= 90 && |
b411b363 PR |
2844 | mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
2845 | dev_info(DEV, "Preparing to skip initial sync\n"); | |
2846 | skip_initial_sync = 1; | |
2847 | } else if (mdev->state.conn != C_STANDALONE) { | |
2848 | retcode = ERR_CONNECTED; | |
2849 | goto out_dec; | |
2850 | } | |
2851 | ||
2852 | drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ | |
2853 | drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */ | |
2854 | ||
2855 | if (args.clear_bm) { | |
20ceb2b2 LE |
2856 | err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, |
2857 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); | |
b411b363 PR |
2858 | if (err) { |
2859 | dev_err(DEV, "Writing bitmap failed with %d\n",err); | |
2860 | retcode = ERR_IO_MD_DISK; | |
2861 | } | |
2862 | if (skip_initial_sync) { | |
2863 | drbd_send_uuids_skip_initial_sync(mdev); | |
2864 | _drbd_uuid_set(mdev, UI_BITMAP, 0); | |
62b0da3a | 2865 | drbd_print_uuids(mdev, "cleared bitmap UUID"); |
87eeee41 | 2866 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2867 | _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
2868 | CS_VERBOSE, NULL); | |
87eeee41 | 2869 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2870 | } |
2871 | } | |
2872 | ||
2873 | drbd_md_sync(mdev); | |
2874 | out_dec: | |
2875 | put_ldev(mdev); | |
2876 | out: | |
8410da8f | 2877 | mutex_unlock(mdev->state_mutex); |
3b98c0c2 LE |
2878 | out_nolock: |
2879 | drbd_adm_finish(info, retcode); | |
774b3055 PR |
2880 | return 0; |
2881 | } | |
2882 | ||
3b98c0c2 LE |
2883 | static enum drbd_ret_code |
2884 | drbd_check_conn_name(const char *name) | |
774b3055 | 2885 | { |
3b98c0c2 LE |
2886 | if (!name || !name[0]) { |
2887 | drbd_msg_put_info("connection name missing"); | |
2888 | return ERR_MANDATORY_TAG; | |
774b3055 | 2889 | } |
3b98c0c2 LE |
2890 | /* if we want to use these in sysfs/configfs/debugfs some day, |
2891 | * we must not allow slashes */ | |
2892 | if (strchr(name, '/')) { | |
2893 | drbd_msg_put_info("invalid connection name"); | |
2894 | return ERR_INVALID_REQUEST; | |
774b3055 | 2895 | } |
3b98c0c2 | 2896 | return NO_ERROR; |
774b3055 PR |
2897 | } |
2898 | ||
3b98c0c2 | 2899 | int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2900 | { |
3b98c0c2 | 2901 | enum drbd_ret_code retcode; |
9f5180e5 | 2902 | |
3b98c0c2 LE |
2903 | retcode = drbd_adm_prepare(skb, info, 0); |
2904 | if (!adm_ctx.reply_skb) | |
2905 | return retcode; | |
2906 | if (retcode != NO_ERROR) | |
2907 | goto out; | |
b411b363 | 2908 | |
3b98c0c2 LE |
2909 | retcode = drbd_check_conn_name(adm_ctx.conn_name); |
2910 | if (retcode != NO_ERROR) | |
2911 | goto out; | |
b411b363 | 2912 | |
3b98c0c2 | 2913 | if (adm_ctx.tconn) { |
38f19616 LE |
2914 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
2915 | retcode = ERR_INVALID_REQUEST; | |
2916 | drbd_msg_put_info("connection exists"); | |
2917 | } | |
2918 | /* else: still NO_ERROR */ | |
3b98c0c2 | 2919 | goto out; |
b411b363 PR |
2920 | } |
2921 | ||
9dc9fbb3 | 2922 | if (!conn_create(adm_ctx.conn_name)) |
b411b363 | 2923 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
2924 | out: |
2925 | drbd_adm_finish(info, retcode); | |
2926 | return 0; | |
b411b363 PR |
2927 | } |
2928 | ||
3b98c0c2 | 2929 | int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2930 | { |
3b98c0c2 LE |
2931 | struct drbd_genlmsghdr *dh = info->userhdr; |
2932 | enum drbd_ret_code retcode; | |
b411b363 | 2933 | |
3b98c0c2 LE |
2934 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
2935 | if (!adm_ctx.reply_skb) | |
2936 | return retcode; | |
2937 | if (retcode != NO_ERROR) | |
2938 | goto out; | |
b411b363 | 2939 | |
3b98c0c2 LE |
2940 | /* FIXME drop minor_count parameter, limit to MINORMASK */ |
2941 | if (dh->minor >= minor_count) { | |
2942 | drbd_msg_put_info("requested minor out of range"); | |
2943 | retcode = ERR_INVALID_REQUEST; | |
2944 | goto out; | |
b411b363 | 2945 | } |
0c8e36d9 | 2946 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
2947 | drbd_msg_put_info("requested volume id out of range"); |
2948 | retcode = ERR_INVALID_REQUEST; | |
2949 | goto out; | |
b411b363 | 2950 | } |
b411b363 | 2951 | |
38f19616 LE |
2952 | /* drbd_adm_prepare made sure already |
2953 | * that mdev->tconn and mdev->vnr match the request. */ | |
2954 | if (adm_ctx.mdev) { | |
2955 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) | |
2956 | retcode = ERR_MINOR_EXISTS; | |
2957 | /* else: still NO_ERROR */ | |
2958 | goto out; | |
2959 | } | |
2960 | ||
d3fcb490 | 2961 | down_write(&drbd_cfg_rwsem); |
3b98c0c2 | 2962 | retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume); |
d3fcb490 | 2963 | up_write(&drbd_cfg_rwsem); |
3b98c0c2 LE |
2964 | out: |
2965 | drbd_adm_finish(info, retcode); | |
2966 | return 0; | |
b411b363 PR |
2967 | } |
2968 | ||
85f75dd7 LE |
2969 | static enum drbd_ret_code adm_delete_minor(struct drbd_conf *mdev) |
2970 | { | |
2971 | if (mdev->state.disk == D_DISKLESS && | |
2972 | /* no need to be mdev->state.conn == C_STANDALONE && | |
2973 | * we may want to delete a minor from a live replication group. | |
2974 | */ | |
2975 | mdev->state.role == R_SECONDARY) { | |
ff370e5a | 2976 | drbd_delete_device(mdev); |
85f75dd7 LE |
2977 | return NO_ERROR; |
2978 | } else | |
2979 | return ERR_MINOR_CONFIGURED; | |
2980 | } | |
2981 | ||
3b98c0c2 | 2982 | int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2983 | { |
3b98c0c2 | 2984 | enum drbd_ret_code retcode; |
b411b363 | 2985 | |
3b98c0c2 LE |
2986 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2987 | if (!adm_ctx.reply_skb) | |
2988 | return retcode; | |
2989 | if (retcode != NO_ERROR) | |
2990 | goto out; | |
b411b363 | 2991 | |
ef356262 | 2992 | down_write(&drbd_cfg_rwsem); |
85f75dd7 | 2993 | retcode = adm_delete_minor(adm_ctx.mdev); |
ef356262 | 2994 | up_write(&drbd_cfg_rwsem); |
85f75dd7 LE |
2995 | out: |
2996 | drbd_adm_finish(info, retcode); | |
2997 | return 0; | |
2998 | } | |
2999 | ||
3000 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) | |
3001 | { | |
f3dfa40a | 3002 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 LE |
3003 | struct drbd_conf *mdev; |
3004 | unsigned i; | |
3005 | ||
3006 | retcode = drbd_adm_prepare(skb, info, 0); | |
3007 | if (!adm_ctx.reply_skb) | |
3008 | return retcode; | |
3009 | if (retcode != NO_ERROR) | |
3010 | goto out; | |
3011 | ||
3012 | if (!adm_ctx.tconn) { | |
3013 | retcode = ERR_CONN_NOT_KNOWN; | |
3014 | goto out; | |
3015 | } | |
3016 | ||
ef356262 | 3017 | down_read(&drbd_cfg_rwsem); |
85f75dd7 LE |
3018 | /* demote */ |
3019 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
3020 | retcode = drbd_set_role(mdev, R_SECONDARY, 0); | |
3021 | if (retcode < SS_SUCCESS) { | |
3022 | drbd_msg_put_info("failed to demote"); | |
3023 | goto out_unlock; | |
3024 | } | |
3025 | } | |
f3dfa40a | 3026 | up_read(&drbd_cfg_rwsem); |
85f75dd7 | 3027 | |
f3dfa40a LE |
3028 | /* disconnect; may stop the receiver; |
3029 | * must not hold the drbd_cfg_rwsem */ | |
3030 | retcode = conn_try_disconnect(adm_ctx.tconn, 0); | |
3031 | if (retcode < SS_SUCCESS) { | |
85f75dd7 | 3032 | drbd_msg_put_info("failed to disconnect"); |
f3dfa40a | 3033 | goto out; |
85f75dd7 LE |
3034 | } |
3035 | ||
f3dfa40a | 3036 | down_read(&drbd_cfg_rwsem); |
85f75dd7 LE |
3037 | /* detach */ |
3038 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
f3dfa40a LE |
3039 | retcode = adm_detach(mdev); |
3040 | if (retcode < SS_SUCCESS) { | |
85f75dd7 LE |
3041 | drbd_msg_put_info("failed to detach"); |
3042 | goto out_unlock; | |
3043 | } | |
3044 | } | |
ef356262 | 3045 | up_read(&drbd_cfg_rwsem); |
85f75dd7 | 3046 | |
f3dfa40a LE |
3047 | /* If we reach this, all volumes (of this tconn) are Secondary, |
3048 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have | |
3049 | * actually stopped, state handling only does drbd_thread_stop_nowait(). | |
3050 | * This needs to be done without holding drbd_cfg_rwsem. */ | |
3051 | drbd_thread_stop(&adm_ctx.tconn->worker); | |
3052 | ||
3053 | /* Now, nothing can fail anymore */ | |
3054 | ||
85f75dd7 | 3055 | /* delete volumes */ |
ef356262 | 3056 | down_write(&drbd_cfg_rwsem); |
85f75dd7 LE |
3057 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { |
3058 | retcode = adm_delete_minor(mdev); | |
3059 | if (retcode != NO_ERROR) { | |
3060 | /* "can not happen" */ | |
3061 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 PR |
3062 | up_write(&drbd_cfg_rwsem); |
3063 | goto out; | |
85f75dd7 LE |
3064 | } |
3065 | } | |
3066 | ||
85f75dd7 LE |
3067 | /* delete connection */ |
3068 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { | |
9dc9fbb3 PR |
3069 | list_del(&adm_ctx.tconn->all_tconn); |
3070 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
3071 | ||
85f75dd7 LE |
3072 | retcode = NO_ERROR; |
3073 | } else { | |
3074 | /* "can not happen" */ | |
3075 | retcode = ERR_CONN_IN_USE; | |
3076 | drbd_msg_put_info("failed to delete connection"); | |
85f75dd7 | 3077 | } |
ef356262 PR |
3078 | up_write(&drbd_cfg_rwsem); |
3079 | goto out; | |
85f75dd7 | 3080 | out_unlock: |
ef356262 | 3081 | up_read(&drbd_cfg_rwsem); |
3b98c0c2 LE |
3082 | out: |
3083 | drbd_adm_finish(info, retcode); | |
3084 | return 0; | |
b411b363 PR |
3085 | } |
3086 | ||
3b98c0c2 | 3087 | int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3088 | { |
3b98c0c2 | 3089 | enum drbd_ret_code retcode; |
b411b363 | 3090 | |
3b98c0c2 LE |
3091 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
3092 | if (!adm_ctx.reply_skb) | |
3093 | return retcode; | |
3094 | if (retcode != NO_ERROR) | |
3095 | goto out; | |
3096 | ||
ef356262 | 3097 | down_write(&drbd_cfg_rwsem); |
3b98c0c2 | 3098 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { |
9dc9fbb3 PR |
3099 | list_del(&adm_ctx.tconn->all_tconn); |
3100 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
3101 | ||
3b98c0c2 LE |
3102 | retcode = NO_ERROR; |
3103 | } else { | |
3104 | retcode = ERR_CONN_IN_USE; | |
b411b363 | 3105 | } |
ef356262 | 3106 | up_write(&drbd_cfg_rwsem); |
b411b363 | 3107 | |
992d6e91 LE |
3108 | if (retcode == NO_ERROR) |
3109 | drbd_thread_stop(&adm_ctx.tconn->worker); | |
3b98c0c2 LE |
3110 | out: |
3111 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3112 | return 0; |
3113 | } | |
3114 | ||
3b98c0c2 | 3115 | void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib) |
b411b363 | 3116 | { |
3b98c0c2 LE |
3117 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3118 | struct sk_buff *msg; | |
3119 | struct drbd_genlmsghdr *d_out; | |
3120 | unsigned seq; | |
3121 | int err = -ENOMEM; | |
3122 | ||
3123 | seq = atomic_inc_return(&drbd_genl_seq); | |
3124 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3125 | if (!msg) | |
3126 | goto failed; | |
3127 | ||
3128 | err = -EMSGSIZE; | |
3129 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3130 | if (!d_out) /* cannot happen, but anyways. */ | |
3131 | goto nla_put_failure; | |
3132 | d_out->minor = mdev_to_minor(mdev); | |
3133 | d_out->ret_code = 0; | |
3134 | ||
3135 | if (nla_put_status_info(msg, mdev, sib)) | |
3136 | goto nla_put_failure; | |
3137 | genlmsg_end(msg, d_out); | |
3138 | err = drbd_genl_multicast_events(msg, 0); | |
3139 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3140 | if (err && err != -ESRCH) | |
3141 | goto failed; | |
b411b363 | 3142 | |
3b98c0c2 | 3143 | return; |
b411b363 | 3144 | |
3b98c0c2 LE |
3145 | nla_put_failure: |
3146 | nlmsg_free(msg); | |
3147 | failed: | |
3148 | dev_err(DEV, "Error %d while broadcasting event. " | |
3149 | "Event seq:%u sib_reason:%u\n", | |
3150 | err, seq, sib->sib_reason); | |
b411b363 | 3151 | } |